Strange Loop Program, Sept 12-14, 2019 - Amazon S3...Open Source is for Everyone 9/12 9:00 am-11:00...
Transcript of Strange Loop Program, Sept 12-14, 2019 - Amazon S3...Open Source is for Everyone 9/12 9:00 am-11:00...
StrangeLoopProgram,Sept12-14,2019HandsonwithDeepLearningandIoT
9/129:00am-11:00am,UnionStationIllinois
Let'sgetstartedwithDeepLearningandtheInternetofThings!We'lldohands-onexercises:you'lluseawebcamandaneuralnetworktorecognizeimages,aggregatedata,andrunreal-timeIoTanalytics.OurgoalistogetyouexcitedaboutIoTandDeepLearning,andtosetyouupforsuccesswithmakerprojectsinyourcommunityaftertheconference.
ShrutiSapre-MathWorks
ShrutiSapreisaSoftwareEngineeratMathWorksandworksintheMATLABLanguagearea.BeforethatshewasintheEngineeringDevelopmentGroupatMathWorkswheresheinteracteddirectlywithcustomerstoresolvetheirtechnicalquestionsonMATLAB.SheholdsaMaster'sdegreeinComputerSciencefromtheUniversityofSouthernCaliforniaandaBachelor'sdegreeinComputerEngineeringfromtheUniversityofPuneinIndia.Herinterestsoutsideofworkincludehikingandtraveling.
MadhumithaRaghu-MathWorks
MadhuisaSeniorSoftwareEngineer@MathWorks.SheworksfortheMATLABObjectSystemteam,whereherfocusinonaddingfeaturestotheMATLABlanguage(specifically,MATLABclasses).Madhuholdsamaster'sdegreeinBiomedicalEngineeringfromCarnegieMellonUniversityandbachelor'sdegreeinBiomedicalEngineeringfromSSNCollegeofEngineering(AffiliatedtoAnnaUniversity).
Sheisintriguedbycomputervisionandmachinelearning.Shehasworkedonafewimageprocessingprojects@SSN,CMUandMathWorks.Inherfreetime,shelearnsIndianclassicalmusicandperformsoccasionally.
ScalableMachineLearningPipelineswithMLflow
9/129:00am-11:00am,UnionStationJeffersonian
DevelopingMachineLearningProductsthatcanscalecomeswithnumerousbarriers.Fromgatheringdata,trainingamodel,anddeployingthemodel,thereisacomplexseriesofstepsneededtobedoneateachsteptoprovidemeaningfuloutput.WithMLflowwecanencapsulatethemachinelearningpipelineakintoaDockerimagetoreducethemanualprocessesneededtodeployamodel.
InthisworkshopwewilluseMLflowtobuildamachinelearningpipeline.WewillexplorehowMLflowallowsustocreatescalable,reproducible,andtrackablemachinelearningpipelinesthroughits3mainmodules.
MLflowTracking-TracksexperimentstorecordandcompareparametersandresultsMLflowProjects-PackagesMLcodeinareusable,reproducibleformtoshareortransfertoproductionMLflowModels-ManagesanddeploysmodelsfromavarietyofMLlibrariestoavarietyofmodelservingandinferenceplatforms.
ThisIsanintermediateworkshopthatwillrequireknowledgeofusingpythonandnavigatingaLinuxterminal.Nomachinelearningexperienceisnecessary.
BanjoObayomi@banjtheman-TwoSixLabs
BanjoisaSeniorResearchEngineeratTwoSixLabs,wherehedevelopsplatformsolutionsforproductizingvariousresearchedbasedprojects.
BanjoreceivedhisB.SinComputerSciencefromUniversityofMarylandCollegeParkin2011,andhisM.SinComputerSciencefromLoyolaUniversityMarylandin2015.BanjoalsoisanAWSCertifiedSolutionsArchitect-AssociateandAWSCertifiedBigDataEngineer.
AnIntrotoProgrammingSoundWithPureData(Pd)
9/129:00am-11:00am,UnionStationMissouriPacific
PureData(Pd)isafree/libredataflowlanguageinsideavisualprogrammingandperformanceenvironmentwithanemphasisonaudio.Usingapatch-cablestylemetaphor(similartoclassicanalogsynthesizers),PureDataprogramsarebuilttocreatecomputermusic,exploresoundart/design,andexperimentwithaudiosynthesistechniques.
Inthishands-onintroductoryworkshop,wewillcoverthefoundationsofprogramminginPdandprovideaverybriefprimeronaudiosynthesistechniques.AttendeeswilllearnhowtocreatePdpatcheswithcommon,first-principlesbuildingblocksandhowtowiretheseobjectstogethertocreateandcontrolsoundsynthesizers.
Somerudimentaryprogrammingexperienceorexposuretosignaltheoryishelpful,butnotrequired.AttendeesshouldbringheadphoneswiththeirLinux,Mac,orWindowslaptop,ideallywithPureData"vanilla"installed.
JasonPlumb@breedx5-gh:breedx2-NewRelic
JasonPlumbisahacker,artist,experimenter,polyglotprogrammer,anddadfromPortland,OR.HewastheDorkbotPDXcommunityoverlordforadecadeandhashelpedtocreatepermanentsoundartinstallationsatawell-knownsoftwarecompany.Hecurrentlyworkswithhigh-volumemetadataatNewRelicandcollaborateswithFutel(toinstallfreepublicpayphones)andtheChurchofRobotron(tosavehumanityfromtheinevitable2084robot/AIapocalypse).
OpenSourceisforEveryone
9/129:00am-11:00am,UnionStationNewYork
We'retoldthatcontributingtoopensourceisgoodforourcareersandagreatwaytolevelupourskills.Let'sfacefacts.Gettingintoopensourceisintimidating,oftenunfriendly,andhardtofindaplacetogetstarted.Cometothisworkshopandlearnthebasicsofcontributingtoopensource.We'lldiscusthingslikeclaimingissues,writinggoodcommitmessages,andfindinggreatprojectstocontributeto.Don'tforgetyourlaptopbecauseeveryattendeewillleavewithapullrequestonanopensourceprojectbeingusedinproductionbyorganizationslikewomen'sshelters,diaperbanks,andotherorganizationsdoingmeaningfulwork!Comelearnnewskillsandmakeadifferenceintheworld!
SeanMarcia@seanmarcia-gh:seanmarcia-RubyforGood
Seanabsolutelylovesthetechcommunity.Heisatirelessdo-gooderwhocreatedandorganizesRubyforGoodandspendshisdayjobworkingtomakegovernmentmoresane.Heiscommittedtomakingtheworldabetterplace.Whennotprogramminghelovesbeingoutdoors(especiallynationalparks),drinkingcoffeefromPortland,eatingdriedseaweedandplayingwithdogs.
ServerlessArchitecturebyExample
9/1212:30pm-2:30pm,UnionStationIllinois
Createaserverlessdistributedprogrammingcompetitionsystemusingavarietyoftechnologies:functions-as-a-service,messagequeues,platform-as-a-service,storage-as-a-service,NoSQLdatabase,event-drivensoftware,andauthentication-as-a-service.ThecompletedsystemisavailableonGithubforattendeestoexamineandmodifyafterseeingthetalk,iftheydesire.ThesystemiswritteninPythonusingGoogleCloudPlatformservices,buttheconceptscanbeadaptedtootherlanguagesandcloudproviders.
Thissolutionconsistsoffourmaincomponents,looselycoupledviaHTTPrequests,messagequeues,andsystemeventtriggers.
ContestparticipantswriteanddeploytheirsolutionsusingCloudFunctionsrespondingtoHTTPrequests.Contestjudgescreatesoftwaretotestthesolutions,alsousingCloudFunctions,buttriggeredbyamessagequeuesothatmultipledifferenttestrunsoneachsubmissioncanrunsimultaneously.JudgingresultsaresavedtoaCloudStorageobject,whichwilltriggeranotherCloudFunctiontoupdateaFirestoreNoSQLdatabasewithresults.AwebapplicationthatletsparticipantssubmittheURLsoftheirsolutionsandviewjudgingresultsisdeployedusingAppEngine,andaccesstoitiscontrolledviaIdentityAwareProxy.
Thetalkstartswiththemotivationforthesystem,whichwastoprovideasimplewayforstudentstosubmitsolutionstoprogrammingcontestproblemsandshowshowthearchitectureevolvedfromthere.
CharlesEngelke@charlesengelke-gh:engelke-Google,LLC
CharlesEngelkeisaSeniorDeveloperProgramsEngineeratGoogleCloud,focusedonapplicationdevelopmentforthecloud.PriortocomingtoGoogle,hewastheCTOofInfoTech,Inc.,whereheledtheefforttobringsealedbiddingtotheInternetandledteamsthatbuildseveralothernative,web,andmobileapplications.
LaurieWhite@lannie29-gh:annie29-Google,LLC
LaurieWhiteisSeniorDeveloperAdvocateatGoogleCloud,focusingonhelpinghighereducationusethecloud,andProfessorEmeritusofComputerScienceatMercerUniversity.Duringhercareerasaprofessorshetaughtawidevarietyofsubjectsincomputerscience,rangingfromintroductoryprogrammingtoprogramminglanguages,softwareengineering,compilerdesign,discretemathematics,andtheoryofcomputation.
ManagingmemorywithRust:dispellingtheFUD
9/1212:30pm-2:30pm,UnionStationJeffersonian
RustisaniftyprogramminglanguagewhichoffersC++-levelperformancewiththesafetyoflanguageslikeJavaandPython.Rustachievesthisbalancingactwithitsunique"borrow-checker",whichverifiesmemorymanagementsafetyatcompiletime.WhiletheRustcommunityandecosystemaregrowingquiterapidly,unfortunately,thereisagreatdealofFUD(Fear,Uncertainty,andDoubt)aroundtheborrowchecker,andmanypeoplesimplydon'ttryRustasaresult.
Inthisworkshop,attendeeswilldevelopacomfortableunderstandingofhowmemorymanagementworksinRust,theroleoftheborrowchecker,andhowtotakeadvantageofRust'scapabilitiestowriteextremelyfastcodethatisimmunetosegfaultsandbufferoverruns.Attendeeswilllearnthat,despitetheFUD,Rust'smemorymanagementisnotmanual.Theywillalsolearnthat,despitetheFUD,theborrowcheckerisyourfriend.
Theworkshopisstructuredasasequenceofshortlessonsalternatingwithhands-onexercises.Bringpenorpencilandalaptopwithanetworkconnection.NopriorRustexperienceisassumed.
JeremyBrown-gh:jhbrown94-Selfemployed
JeremyBrownreceivedhisPhDincomputerscienceandelectricalengineeringfromMITin2002,afterwhichheworkedatBluefinRoboticsandITASoftwarebeforeco-foundingJaybridgeRoboticsin2008.After8yearsofmakingtractorsandminingtrucksintoautonomousvehicles,theJaybridgeteammovedenmassetoToyotaResearchInstitutein2016.JeremyremainedatTRIuntil2018.Heispresentlyself-employedasanentrepreneur-on-break,whichisafancywaytosaydeliberatelyunemployed.Heispresentlyworkingtosharpenhistechnicalskills(whichhadbeenlightlydulledbyyearsasafull-timemanager),todevelopnewbusinessplans,and(soon)toseekco-foundersforhisnextstartup.JeremylivesinSomerville,Massachusetts.HeismarriedtoProf.DanieleLantagne.Theyareraisingtwosmallchildrentogether.
DayofDatomicCloud
9/1212:30pm-2:30pm,UnionStationMissouriPacific
DatomicCloudÂisatransactionaldatabasewithaflexibledatamodel,elasticscaling,andrichqueries.Datomicletsyoufocusonyourapplicationlogic,writingordinaryClojurefunctionswithin-memoryaccesstoyourdata.Datomichandlesthedeploymentandexecutiondetails.
Inthis4hrworkshop,youwilllearnabout:
Datomic'sinformationmodel,schema,ACIDtransactions,anddatalogquerylanguageIonsÂtodeployentireClojureapplicationstoDatomicImplementingwebserviceswithIons

MarshallThompson@GlassOnion9-gh:glassonion-Cognitect
I'mamulti-disciplinaryscientistandsoftwareengineerinterestedinapplyingcutting-edgesoftwarepracticestogenomic-scalebiologicalproblems.Icaredeeplyaboutscienceandengineeringeducationandgreatlyenjoyteaching.Ialsolikemotorcycles.
TechnicalZineWritingWorkshop
9/1212:30pm-2:30pm,UnionStationNewYork
Thisworkshopwillbeanintroductiontotechnicalzinemakingandwriting.Participantswillexplorehowtoexplaintechnicaltopicsviadrawings,comics,diagrams,andstories,withparticularfocusonmakingcomplextopicsunderstandableandaccessible.Attheendoftheworkshop,participantswillhavetechnicalzinesthattheycanexchangewitheachotherandotherconferenceattendees.
AmyWibowo@sailorhg-gh:sailorhg-BubblesortZines
isputtingthe❀soft❀insoftwareengineeringbywritingzinesaboutalgorithms,encryption,anddatastructureswithlotsofillustrations,buildingpastelhardware,anddesigningtech-themedaccessoriesandapparelathercompanyBubblesortZines.
CodeThat(Practically)WritesItselfWithTypeScript
9/123:00pm-5:00pm,UnionStationIllinois
JavaScriptisapowerfullanguagethat'sflexible-toafault.TypeScriptaddsalayerofstatictypingovermodernJavaScriptthatmakescodestronger,morerobust,andeasiertowrite.Inthisworkshop,we'regoingtodiveintoTypeScriptfundamentals,type-firstdevelopmentpatterns,andahandfulofmagictricks!
Requirements:YourownWindowsorMaclaptop.DownloadrequirementsandsetupinstructionscanbefoundintheREADMEathttps://github.com/atomicobject/ts-workshop.
RachaelMcQuater@excitedNerd-gh:rachaelmcq-AtomicObject
DrewColthorpisaprincipalconsultantandRachaelMcQuaterisadeveloperatAtomicObject.DrewhasbeenseekingoutbetterwaystowriteJavaScriptfor20years,andRachaelisjustgettingstartedwiththesame.We'rebothreallyexcitedaboutthepowerandbeautyofsoftwarethatcloselymodelscomplexreal-worldproblemdomains.
DrewColthorp@dcolthorp-gh:dcolthorp-AtomicObject
DrewColthorpisaprincipalconsultantandRachaelMcQuaterisadeveloperatAtomicObject.DrewhasbeenseekingoutbetterwaystowriteJavaScriptfor20years,andRachaelisjustgettingstartedwiththesame.We'rebothreallyexcitedaboutthepowerandbeautyofsoftwarethatcloselymodelscomplexreal-worldproblemdomains.
HowtoLeverageYourOwnTalentbyMentoring
9/123:00pm-5:00pm,UnionStationJeffersonian
Youareatop-notchdeveloper.Highlysoughtafterandveryeffective.Howcanyouleverageyourtalentandinfluencewithouttravelingdownthemanagementpath?BecomeaMentor.Thisworkshopwillcoverthe5essentialskillsyou'llneedinordertomentor:MasteringFlow,GivingFeedbackandGuidance,MentoringwithMultipleIntelligences,SettingClearGoalsandIndicators,andBuildingConfidence.Wewillusehands-onjugglingexercisesasaplaygroundtopracticetheseskills.
DougBradbury@dougbradbury-gh:dougbradbury-8thLight
Dougwasanearlycontributorto8thLightwhenhejoinedin2007.DoughelpedstarttheChicagoofficein2010andhasperformedalmosteveryroleasacrafterinthattime.As8thLighthasgrown,Doughasdefinedandthenservedinmanynewroles.IncludingDirectorofConsulting,DirectorofStudio,ManagingDirector,andCOO.Dougisnowresponsibleforthementoringprogramat8thLight.DougearnedhisBachelorofSciencewithHonorsinComputerEngineeringfromtheUniversityofIllinoisin2002.Hehasmentoredmorethan40of8thLight'ssoftwarecraftersandhasgivenseveralpresentationsonapprenticeshipsinsoftware,acceptancetesting,Agilesoftware,andmentoring.HeauthoredtheManifestoforSoftwareCraftsmanshipandisapart-ownerof8thLight.
CloudNativeKubernetesforNode.jsDevelopers
9/123:00pm-5:00pm,UnionStationNewYork
Ascloudservicesanddeploymentplatformshavecometobearmoreandmoreresponsibilityforthe"Enterprise"featuresofmodernapplicationdevelopment,Node.jsbecomesmoreappealingfordeveloperswhowanttotakeadvantageofitslightweight,quickstartuptime,developerproductivityandoverallhappiness.Withcloudnativeapplicationdevelopmentpatterns,supportedbyaKubernetesdeploymentplatform,Node.jsisaformidableenterpriseplayer.
Butcloudnativeapplicationdevelopmentcanseemdaunting.HowdoyoubothwriteanapplicationandalsodealwithDevOpsautomation,servicediscovery,observability,imageimmutability,scalability,failoverandalloftheothercloudnativeandenterpriseideals?
It'salottokeeptrackof.It'salottodo.It'sScary!
Butitdoesn'thavetobe.InthistalkIwillshowyouhowtomakeKubernetesNode.jsapplicationdeploymentsimpleandstraightforward.Takeajourneywithmetothecloud.Bythetimewe'redone,attendeeswillhavecreatedaNode.jswebappanddeployedittoaKubernetescluster.
LanceBall@lanceball-gh:lance-RedHat
LanceisaPrincipalSoftwareEngineeratRedHat,leadingtheNode.jsengineeringeffortswithinRedHatMiddleware.Heisanactiveopensourcecontributor,anauthorofseveralNPMmodules,andacorecommitterfortheNode.jsproject.InadditiontoNode.js,you'llfindhiscontributionsinmanyotherwell-knownopensourceprojectsincludingWildflySwarm,Vert.xandTorquebox.
StrangeLoopParty
9/127:00pm-9:30pm,CityMuseum
We'retakingovertheCityMuseumfortheStrangeLoopparty!TheCityMuseumisawarehousepackedfullofexplorableartmadefromrepurposedmaterials.Themuseumisfractalinnatureanddifficulttodescribe.Itfeaturesallmannerofslides,tunnels,and
beautifulweirdness.Theyhaveastrict"NoMaps!"policy-it'suptoyoutofindyourownpath.
Shuttles
Shuttleswillbeavailablecontinuouslygoingto/fromCityMuseumtoUnionStation(20thStreet),HiltonBallpark,andPearTreeInn.
Howtoteachprogramming(andotherthings)?
9/139:10am-10:00am,StifelTheatreTheater
Everyoneshouldlearnprogramming,right?Yes!Buthow...Shouldweallowchildrentoexploreandlearnaboutsyntaxontheirown,orshouldwedrillprogramminglikewerotememorizethetableofmultiplicationorGermangrammaticalcases?Felienne'stalkoutlinesthishistoryofprogrammingeducationanddidacticsbeliefsinprogrammingthatleadtotheprevalenceofexploratoryformsofteaching,startingwithPapert'sLOGO.Shewillthenexploreprogrammingeducationinrelationtomathematicsandlanguageeducationandexplorehowrotelearningcouldlooklikeforprogramming.Feliennewilldiscussherownresearchintomisconceptionsandcodephonologyasmeanstoteachprogrammingmoreeffectively.
FelienneHermans@Felienne-gh:Felienne-LeidenUniversity
FelienneisassociateprofessorattheLeidenInstituteofAdvancedComputerScienceatLeidenUniversity,wheresheheadstheProgrammingEducationResearchLab(PERL).Felienne'sworkfocusesonthequestionhowtobestteachprogrammingtokidsandstudents,investigatingmisconceptions,didacticstrategiesandprogrammingenvironments.Felienneisaco-editoroftheACMjournalTransactionsonComputingEducation,andahostattheIEEEpodcastSERadio,oneofthemostpopularsoftwareengineeringpodcastsontheweb.ShewritesamonthlycolumnforDutchITmagazineAGConnectandsherantheJoyofCodingconferencefor6years.Whensheisnotcoding,bloggingorteaching,sheisprobablyknitting,runningorplayinga(board)game.
Felienneblogsatfelienne.com
TypingtheUntyped:SoundnessinGradualTypeSystems
9/1310:20am-11:00am,StifelTheatreEncore
Recentyearshaveseenanexplosionofgradualtypesystemsandsupersetlanguagesthataddtypestopreviouslyuntypedlanguages:TypeScript&FlowforJavascript,MyPyandPyreforPython,HackandPHP7forPHP,SorbetforRuby,andmanymore.Implementingthesetypesystemsinvolvesmakingtradeoffsbetweensoundness(catchingasmanyerrorsaspossible)andcompleteness(notrejectingvalidprograms)thatfundamentallyimpacttheusabilityandusefulnessofthetypesystem.
Inthistalk,I'llexamineafewofthesetradeoffsthatapplyacrossmanylanguages:typerefinementandrefinementinvalidation,arrayout-of-bounderrors,andvariance(particularlyarraycovariance).We'lllookatwhattradeoffsagradualtypesystemneedstomake,whattheadvantagesaretodifferentapproaches,andcomparehowvariousgradualtypesystemsandmoretraditionalstatictypesystemshandlethesetradeoffs.
BenWeissmann-gh:benweissmann-Tulip
Ben"Fuzzy"WeissmannisasoftwareengineeratTulip,wherehewasthefirstemployeeandleadsarchitectureonthePlatformteam,creatingaplatformformanufacturerstobuildappsthatstreamlinetheiroperations.Hisfocusisonarchitecture,backendsystems,anddevelopertooling.Inthepast,he'sworkedatTwitter,TripAdvisor,andtheMITMediaLab.
Recreatingforgottenprogramminglanguages,forart!
9/1310:20am-11:00am,StifelTheatreOvation
Theearlybeginningsofcomputergraphicsinthe1960ssawthebirthofanumberprogramminglanguagesthatwerecreatedspecificallyformakinganimationsandgraphics.Almostallofthemarenowobsoleteandmostlyforgotten.However,backthen,manyoftheselanguagescutting-edgeandmadepossiblethecompletelynewfieldofmakingartwithcomputers.AprolificexampleofthiswasBellLabs'BEFLIX,alanguagecreatedtomakeanimationsusingamicrofilmplotter.
Survivingdocumentationoftheselanguagesarescarce,muchlesstheactualcompilersorinterpretersthemselves.Theyweresimplelanguages,yettheywereusedtocreatesomeofthemostcomplexandcompellinggraphicsevenbytoday'sstandard.Inordertounderstandhowtheyweremade,aswellasthelimitationsofcomputergraphicsthatartistsandprogrammerswereworkingwith,IattemptedtorecreateaninterpreterforART1.ART1wasunique,itwasthefirsteverdocumentedprogramminglanguagedesignedspecificallyforthosenewtoprogrammingtocreateart.Inthistalkwe'llseehowIhuntedforsurvivingdocumentation,builttheinterpreterandfinallyusedittoreverseengineersomeoftheearliesticoniccomputerartpieces.Expecttoseelotsofvintagecomputerart!
SherMinnChong@piratefsh-gh:piratefsh
SherMinnisawebplatformengineerbydayandacasualgenerativeartistatnight.Someofherfavoritethingsarewebtech,drawingcomics,andteachingprogramming.
DiveintoStreamswithBrooklin
9/1310:20am-11:00am,StifelTheatrePeabodyLounge
AlthoughLinkedIndatacontinuestogrowrapidlyovertheyears,scalinguptohandletheincreasingdatavolumehasnotbeentheonlychallengeinstreamingdatainnearreal-time.SupportingtheproliferationofnewdatasystemshasbecomeyetanotherhugeendeavorfordatastreaminginfrastructureatLinkedIn.Buildingseparate,specializedsolutionstomovedataacrossheterogeneoussystemsisnotsustainable,asitslowsdowndevelopmentandmakestheinfrastructureunmanageable.Thiscalledforacentralized,managed,andextensiblesolutionthatcancontinuouslydeliverdatatonearlineapplications.
WebuiltBrooklinasamanageddatastreamingservicethatsupportsmultiplepluggablesourcesanddestinations,whichcanbedatastoresormessagingsystems.Since2016,BrooklinhasbeenrunninginproductionasacriticalpieceofLinkedIn'sstreaminginfrastructure,supportingavarietyofdatamovementusecases,suchaschangedatacapture(CDC)anddatapropagationbetweendifferentsystemsandenvironments.WehavealsoleveragedBrooklinformirroringKafkadata,replacingKafkaMirrorMakeratLinkedIn.Inthistalk,wewilldivedeeperintoBrooklin'sarchitectureandusecases,aswellasourfutureplans.
CeliaKung-LinkedIn
CeliamanagesthedatapipelinesteamatLinkedIn.Previously,shewastheleadengineerforbuildingOraclechange-datacapturesupportforBrooklin,aswellasanewKafkamirroringsolutionthathasfullyreplacedKafkaMirrorMakeratLinkedIn.
HowtoFixAI:SolutionstoMLBias(AndWhyTheyDon'tMatter)
9/1310:20am-11:00am,StifelTheatrePremier
BiasinmachinelearningisaProblem.Thisiscommonknowledgeformanyofusnow,andyetouralgorithmscontinuetooperateunfairlyintherealworld,perpetuatingstructuralinequalityalonglinesofclassandcolor.Afterall,"bettertrainingdata"isnotsoeasytogetourhandson,right?
Inthistalk,Iarguethatitistimeforustobeginbuildingalgorithmsthataredesignedtoberesilienttobiaseddata.BuildingonabasicintroductiontoMLconcepts,Ipresentanin-depth,intuitiveexplanationofseveraldeeplearningtechniquesthatcombatunderlyingbiasindata,andusethesemodelstoexplorewhat"algorithmicfairness"reallymeansinmeasurableterms.Finally,divingintoafewcasestudiesofrealworldsystems,Isuggestthatevenperfect"fairness"isnotnecessarilythefairy-taleendingweliketothinkitis.BlindlyoptimizingforitmaystillmisstherealproblembehindAIbias,andtocometoarealsolutionwemayjusthavetoreframetheproblemitself.
JoyceXu@joycex99-gh:joycex99-SidewalkLabs
JoyceisanAI/MLengineer(tinkerer?)whomightbealittletooexcitedabouthistory,urbanstudies,andbingingHBOtobeaproductivetechworker.SheiscurrentlyatSidewalkLabs,whereshe'sthinkingabouthowtoengineerprivacy-preservingMLsolutionsinurbanmobilityandsustainability.Previously,sheconductedresearchatDeepMindandtheStanfordNLPGroup,whereherpursuitscenteredaroundmulti-agentreinforcementlearningandnaturallanguagegenerationrespectively.HavingbegunherjourneyinAIself-taught,sheisastrongadvocateofaccessibilityinresearchandtooling:she'shelpedbuildanopen-sourceMLframeworkforfunctionalprogramminginClojure,andblogsregularlyonAIresearchandfundamentals.Whenshe'snotbusydaydreamingaboutbecomingaDJ,shesometimeslooksforwardtofinishingheruniversitystudiesatStanford.
BehindTheNewYorkTimesCrossword
9/1310:20am-11:00am,StifelTheatreTheater
TheNewYorkTimesCrosswordisseriousbusiness.Operatedlikeastartupwithinthecompany,theNYTGamesTeamworkstirelesslytokeepour500,000+subscribershappilysolvingwhiledrivingsignificantrevenuetosupportgreatjournalism.Withserioussolverscomestheneedforserioustechnology.SincerewritingtheentireplatforminGoandmigratingittoaserverlessecosystem,theGamesTeamhasdriveninnovationatTheTimeswiththemostcuttingedgetoolsandinfrastructure.Takeabehind-the-sceneslookofhowapuzzlemakesitfromaconstructortooursolversaroundtheworldandthetechchallengeswe'vehadtoovercometoenableit.Seehowwe'vebuiltourgamesplatformtohandleimmensetrafficspikeswhennewpuzzlesarereleased,allowustofearlesslytestinandpushchangestoproduction,andquicklylaunchnewgamesforoursubscriberstopuzzleover.
DarrenMcCleary@darren_out-TheNewYorkTimes
I'maSeniorSoftwareEngineeratTheNewYorkTimes.I'vebeenamemberoftheNYTGamesTeamsince2016.I'mamemberofthe
Times'ArchitectureReviewBoardwhereIguidetechnologydecisionsandpracticefortheentirecompany.
MakingsofaModernACIDCompliantDistributedDatabase
9/1310:20am-11:00am,UnionStationGrandF
There'splentyofdistributeddatabasesonthemarketthesedays.There'sonlyafewthatarebothdistributedandprovideACIDguarantees.Inthistalk,I'llbesheddinglightonsomeofthemoreinterestingaspectsoftheinternaldesignofFaunaDB,onesuchdatabasethatIalsohappentoworkon.We'lllookatCalvin,theprotocolthatensuresdeterministictransactionapplicationinadistributedenvironmentwithverylittlecoordination(muchlessthantwo-orthree-phasecommitprotocols.)We'llseehowwecanscaleCalvinforthroughput.We'lllookathowyoucanmakeadistributedtransactionsystemindependentofclocks.(It'snevergoodtohaveadistributedsystemdependonclocksforcorrectness,yetmanydatabasesdo.Thisonedoesnot.)We'lltalkabouthowyoucantakeacorrectlyoperatingdatabaseandmakeitfasttoowithjudiciousapplicationofcontroltheory:intowhatpointsofthesystemcanyouplugineitheraPIDcontrolleroraKálmánfiltertofurtherreducelatencies.Welearned(andarestilllearning)alotwhileimplementingFaunaDB,andinthistalk,wewanttosharesomeofwhatwelearnedsofarwithyou!
AttilaSzegedi@asz-gh:szegedi-FaunaInc.
AttilaSzegediismostnotedasanalumnusofbothTwitterandOracle'sJavaLangToolsgroup.AtTwitter,heworkedonTwitter'slarge-scaleinfrastructure'sJVMperformanceproblems.AtOracle,hewasdrivingthedevelopmentofthejdk.dynalinkpackageandwaspartoftheteamthatcreatedtheNashornJavaScriptruntime.Outsideofpaidwork,AttilausedtobeamajorcontributortoopensourceprojectssuchasApacheFreeMarkerandMozillaRhino.Somehoweveryprojecthetakesuponendsupbeingeitheradistributedsystemoralanguageruntime.WithhimcurrentlyworkingonmanyaspectsofFaunaDB,hemight'vefinallystumbledacrossaprojectthatisbothofthesethingsatthesametime.AttilaisaJavaChampion,atwo-timeJavaOneRockStar,andthiswillbehisthirdtimehavingthehonorofspeakingatStrangeLoop.
BetterSpotifyPlayliststhroughDiscreteOptimization
9/1310:20am-11:00am,UnionStationRegencyAB
IamaSpotifyaddict,formerDJ,amateurmusician,andprofessionalsoftwareengineer.Itakespecialprideinmakingexpertly-curatedplaylistsformyselfandfriends.Ittakesalotoftimeandenergytosettherightmoodandtone,andevenmoretimeandenergytotransitionsmoothlyfromonesongtoanotherinawaythatmakessenseandispleasingtotheear.Throughmanyyearsofpractice,I'veobservedthatmakingagoodplaylistisalotlikesolvingapuzzle;andjustlikepuzzles,therearerulesandpatternstofollowifyouwanttoproduceacohesiveoutput.Inthistalk,we'llexplorethenotionofteachingtheserulestoacomputer,buildingaplanning&optimizationalgorithmthatfollowstheserules,andlettingitlooseonasetoftrackstogeneratedelightfulplaylistsonSpotify.We'llalsocoverthebasicsofmusictheoryandwhycertainsongssoundbettertogether.Therewilllikelyalsobefasttalking,livekeyboardplaying,andsomeunrehearseddemosagainstarandomsampleofSpotifyplaylistssubmittedbytheaudience.
CedricHurst@divideby0-gh:divideby0-SpantreeTechnologyGroup,LLC
CedricistheFounder&PrincipalofSpantree,aboutiquesoftware+data+systemsengineeringfirminChicago'sWestLoop.Someonerecentlydescribedhimasbeing"likeHouseMDbutfortechnicalstuff."Whileitwasaflatteringcomplimentinsomerespects,hehopesthecomparisondoesn'textendtohisbedsidemanner.He'spassionateaboutdrivingtechnicalchangewithinhisownteamandwithhisclients,buthealsorealizesthatsuccessfultechnicalchangefrequentlynecessitatesculturalchangeaswell.Tothatend,heandhisteamfocusondefininglanguageprotocolsandframeworksthatempowertheteamstheyworkwithtomakeinclusiveandprincipleddecisionsaboutthebestpathforwardtogether.Ifyoulethim,he'lltalkyourearoffaboutdistributedsystems,rulesengines,andexoticdatabases;buthealsoenjoysmakingcomputermusic,travelingandgivingawaymoneytostrangerswithcoolideas.
Empoweringpeopletobuildadigitalbank
9/1310:20am-11:00am,UnionStationRegencyC
Wearegoingtotalkabouthowwebuilt,maintainandscaleourmicroservicesarchitecture.Weareafast-growingdigitalbankwithalotofchallengesregardingscalability,operabilityandreliabilityandwewouldliketopresentastructuredtalkaboutthekeycomponentsofourecosystemandalsohowtheyinteractwitheachother.Sinceday-onewearchitecturedourentireinfrastructuretorunoncloudandtobeplatformagnostic,thatgaveustheabilityofgrowingfastandensuringourhighstandardsreliability.NowadayswedecoupleourinfrastructurefromEC2AmazoninstancestoKuberneteswherewegainedthedesiredagnosticism.There'safewcharacteristicsthatarenotsousualandgivesusalotofleveragewhencomparedtoourcompetitors,thingslikeShardingandHomogeneousCodebase,usingfunctionalprogrammingasourmainparadigm.Also,webuildandmaintainabstractionsthathelpourengineeringteamtosmoothlyoperateandconstantlyimprovetheirmicroservices,andfurthermoretheproductsweoffertocustomers.OneofthingsweusetoachievehighlevelsofresilienceandreliabilityisKuberneteswithourowndevelopedtoolingandabstractions,whichallowsustoprovidefastinteractioncycleandanevenmoreoptimizedwaytodeployservicesandotherinfrastructureparts.Forinstance,wedeploythingslikeThanos,PrometheusandGrafanausingthoseabstractionstoensurefast-ishmannertoreleaseandcontrolsoftwareversioning.
EricYoshimura@ericyoshimura_-gh:ericyoshimura-Nubank
EricYoshimuraisaSoftwareEnginneratNubank,currentlylivinginSãoPaulo,Brazil.NowadaysworkingonaSREteamhelpingtobuildanawesomemonitoringplatform.Insparetime,EriclikestowatchNHLgamesandlearnabouteconomicsstuff.
RiadVargas@riadvargas-gh:riadvargas-Nubank
RiadVargasisaSoftwareEngineeratNubank,withmorethan2yearsofexperiencedevelopingandmaintainingmicroservicesinthefinancialindustry,focusedinfunctionalprogramming,transactionaldatabasesandSREstuff.
TemporalDatabasesforStreamingArchitectures
9/1311:20am-12:00pm,StifelTheatreEncore
Timeisintrinsictoinformationandyetitisusuallyanafterthoughtindatabasedesigns.WepresentCrux,ageneralpurposeopensourcedocumentdatabasewithbitemporalgraphqueries.
ThistalkwillexplorethejourneyofhowCruxwasconceivedfromJUXT'sconsultingexperiencesofbuildingglobalsystemintegrationsandprovidingtemporalquerycapabilitiesforfinancialservicesprojects.
JonandJeremyfromthecoreteamwilloutlinetheoverarchinggoalsforCruxasan"unbundled",pluggable,andlayereddatabasewithbitemporalACIDtransactions.Bitemporalitycleanlyseparatesthetimewhenthesystemlearnsaboutafact(transactiontime)fromthetimesrelatingtotheactuallife-cycleofafact(validtime).Modellingvalidtimeisessentialinastreamingarchitectureformaintainingaconsistenttemporalviewofdata.BitemporalityinCruxprovidesefficientandhorizontallyscalablepoint-in-timeDatalogquerieswhilstenablingprogrammaticuseofthedatabase"asavalue"withgloballyconsistentqueries.
TheteamwillalsodiscussthedesignchoicesmadetosupportaDatalogqueryengineon-topofKey/ValuestoressuchasRocksDBandLMDB,theuseofsemi-immutablelogstoragetosupportevictionofsensitivedata,clusteringwithKafka,andtheuseofClojureforconstructingalayeredandextensiblesystem.Finally,theywillshareaviewonfuturedirectionsandhowJUXTplanstocollaboratewiththeopensourcecommunity.
JeremyTaylor@refset-gh:refset-JUXTLTD
JeremyTaylorisanOfferingManageratJUXTwhereheleadsthedevelopmentofCruxasanopensourcedatabaseproduct.Following7yearsoftechnicalpre-salesforIBM'seCommercesystemsbusiness,JeremytransformedhisinterestinClojureandDatalogfromahobbyintoacareerevolutionbyjoiningJUXT,theUK'sforemostClojureconsultancy.AsanOfferingManager,JeremyisresponsibleforleadingthegrowthofCruxtowardslong-termviabilityasasupportedopensourceproductintheenterprisedatabasemarketplace.Jeremyispassionateaboutdatabasetechnologiesandtheiruseforknowledgerepresentation,andhehascontributedtotheDataScriptlibrarythroughthecourseofhispersonalresearch.
JonPither@jonpither-gh:jonpither-JUXTLTD
JonPitherisatechleadbytradehavingledprojectsforJUXTatmultipleTier-1InvestmentBanks,anonlinenewspaperwebsite,amajorpropertyportal,andaninternationalpublicelectricbikesscheme.Jonisaco-founderofJUXT;asoftwareengineeringfirmthatusesClojureasitsprimaryweaponofchoice.Beforeco-foundingJUXTJonworkedasaseniorconsultantatThoughtWorkswherehebecamewellversedinAgilesoftwaredevelopmenttechniques.DuringhistimeasaconsultantJonworkedatlargeinstitutionssuchastheBBCandSky,amajorpublishinghouseandafashionretailgiant.JonwritesaseriesofblogarticlesforJUXTincludingaseriesofÂreferenceClojureadoptionstories.
ARobotPoetGoesforaWalkinthePark
9/1311:20am-12:00pm,StifelTheatreOvation
Designingcomplex,dynamicsystemsthatcanproduceinterestingandaestheticallypleasingartisaveryhardproblemtosolve,evenwhenyou'rejusttalkingaboutsomethingasfocusedasaTwitterbot.Whathappenswhenyoutrytomakeprocgenartthatdoesn'tjustexistontheInternet,butactuallylivesinandinteractswiththerealworld?
ComputationalFlâneurisagenerativesite-specificpoetrywalkthatcameoutofmyresearchattheMITMediaLab.AsyouwanderthroughtheFortMasonParkinSanFrancisco,apoetryrobotmakesupandreadsyoupoemsbasedonwhereyouwalk.Walkbythebay,andyou'llhearpoemsofthesea;walkbythecannonsandyou'llhearpoemsofwar.
ThistalkwillexplorethedesignanddevelopmentofComputationalFlâneur.Onatechnicallevel,we'lldiscusstheneuralnetworkmodelthatpowersthepiece'spoetrygenerator,includingabriefoverviewofcommontextgenerationtoolsandhowthisexactstructurewaschosentocreateaspecific"bot-like"aesthetic.We'llalsodiscusshowit'sseededbyvariousformsofsmartphonesensordatalikelocationandweather,whatsortsofprototypingprocesseswereneededtofine-tuneit,andhowtoconceptuallyapproachthissortofreal-
worldprocgendifferentlyfromnormalproceduralgenerationproblems.
EmLazer-Walker@lazerwalker-gh:lazerwalker-Microsoft
EmisaToronto-basedartist/engineerwhomakesinteractiveart,experimentalgames,andsoftwaretools.Mostofherworkfocusesonusingnontraditionalinterfacestoreframeeverydayobjectsandspacesasplayfulexperiences.She'sbuiltprojectsasfar-flungasasite-specificgenerativepoetrywalk,agameplayedon19thcenturytelegraphhardware,andacommercialboardgamethatusesAmazonAlexa.Currently,sheworksasaCloudAdvocateforMicrosoftAzure.
EasyAbstractInterpretationwithSPARTA
9/1311:20am-12:00pm,StifelTheatrePeabodyLounge
Usingabstractinterpretationtobuildascalabletoolfromscratchisadauntingengineeringtaskthatgenerallyrequiresaprotracteddevelopmenteffort.Tostreamlinethatprocess,webuilt(SPARTA)[1],aC++libraryofcomponentsforbuildinghigh-performancestaticanalyzersthatcanruninaproductionenvironment.SPARTAprovidesthebuildingblockssoanengineercanfocussolelyonthelogicthatextractsthedesiredinformationfromtheprogram.Thelibraryensuresthattheresultinganalysisissound,efficient,andscalable.
SPARTAisactivelyusedbyFacebook,mostnotablyinouropen-sourceAndroidbytecodeoptimizingcompilerRedex.However,SPARTAislanguage-independentandcontainsnoAndroid-specificlogic.
InthistalkwewillgooversomeofthealgorithmsbehindSPARTA,thebenefitswe'veobtained,andwewillshowyouhowtouseitinyourownanalyses.
[1]:https://github.com/facebookincubator/SPARTA
ArnaudVenet-gh:arnaudvenet-Facebook
ArnaudgothisPhDinabstractinterpretationfromtheÉcolePolytechniquein1998andhashelpedspreadtheuseofscalablestaticanalysisinindustrysincethen.
JezNg-gh:int3-Facebook
JezhasbeenhackingonprogramminglanguagesandruntimesatFacebookforthepastfiveyears:firstontheHacktypecheckerandthenontheRedexoptimizingcompilerforAndroid.WhileanundergraduateatAmherstCollege,healsoworkedonDoppio,oneofthefirstJVMimplementationsinthebrowser.
RubyontheAppleII:AdventuresinRetroProgramming
9/1311:20am-12:00pm,StifelTheatrePremier
TheminimumsystemrequirementsforPowerPoint2019are4GBofRAManda1.6GHzCPU.ThesystemrequirementsforthistalkareafewkilobytesofRAM,a51/4"floppydiskdrive,anda30yearoldcomputer.
Withan8-bitCPUrunningat1megahertzandkilobytesofRAM,theApple][wasahumblemachine.Manypeoplefirstlearnedtoprogramonit,butthelanguagesdesignedfortheApple][tendedtobeclumsyandlimitingbymodernstandards.Surelyitisn'tpossibletofitalanguageasdynamic,rich,andcomplicatedasRubyonsuchalimitedmachine,right?
ComeseeRubyrunningwhereithasneverrunbefore,learnhowprogrammingwasdonedecadesago.
ColinFulton@PeterQuines-gh:justcolin-DuoSecurity
Comingfromtheworldoftheatredesignandart,Colinisafront-enddeveloperandwebaccessibilityspecialistatDuoSecurity,adivisionofCisco.Whentheyaren'tatworktheyaredoingsillythingslikeimplementingchessgamesinuntypedlambdacalculusorwritingConway'sGameofLifeinsignificantlyfewercharactersthanittooktowritethisbio.
Observability:SuperpowersforDevelopers
9/1311:20am-12:00pm,StifelTheatreTheater
Whowantstospendtimedreamingabouttheabilitytoleaptallbuildingswithasinglebound,whenwecanrecaststorieswelivedaytodayaspowersofourown...andimproveourownlivesinapracticalway?Whenobservabilityisfoldedintothedevelopmentprocessitself,itrepresentsthepotentialforabeautifullyvirtuouscycle:productionstopsbeingjustwhereourdevelopmentcoderunsintoissues,
anditbecomeswherepartofourdevelopmentprocesslives.
ChristineYen@cyen-gh:christineyen-Honeycomb
Christinedelightsinbeingadeveloperinaroomfullofopsfolks.AsacofounderofHoneycomb.io,astartupwithanewapproachtodebuggingproductionsystems,shecaresdeeplyaboutbridgingthegapbetweendevsandopswithtechnologicalandculturalimprovements.BeforeHoneycomb,shebuiltoutananalyticsproductatParse(boughtbyFacebook)andwrotesoftwareatafewnow-defunctstartups.
LearningtoLoveJavaScript
9/1311:20am-12:00pm,UnionStationGrandF
JavaScriptisanimperfectprogramminglanguage.It'sweakly-typed,scopingrulesandtypecoercioncanmakedifficult-to-diagnosebugs,andcross-browsercompatibilitysometimesfeelslikeapipedream.Butdoesitmatter?
JavaScriptisthemostpopularprogramminglanguageintheworld,andcontinuestogrowyearafteryear.Howdidthathappen?HasJavaScriptsucceededdespiteitswarts,ordiditsshortcomingshelppavetheroadtoitswidespreadadoption?
Inthistalk,I'llarguethatjudgingthemeritsofJavaScriptsolelyasaprogramminglanguageismissingthepoint,andtoaccuratelyweighitsplaceintheworld,youmustconsiderJavaScriptbeyondtypes,scope,andsyntax.You'llwalkawaywithadeeperappreciationforthefeatofhumancooperationthatistheWorldWideWeb,andhowJavaScriptbecameoneoftheWeb'smostimportantthreads.
JavaScriptisheretostay.Let'slearntoloveittogether.
TaraVancil@taravancil-gh:taravancil-Glitch
TaraisadeveloperatGlitch,thefriendlycommunitywhereanyonecanbuildwebsites,apps,bots,andmore.Previously,sheco-foundedtheBeakerBrowser,anexperimentalWebbrowserthatexploreshowpeer-to-peerprotocolscanmakebuildingandsharingontheWebmoreaccessible.HerworkisrootedinunderstandingtheWebasauniversalcanvasforcreationandexpression,butalsoasanevolvingplatformthatmustalwaysbecarefullycultivated.ShelivesinAustin,TX.
RGBtoXYZ:TheScienceandHistoryofColor
9/1311:20am-12:00pm,UnionStationRegencyAB
Colorisafascinatingsubject.Itisbothincrediblysubjective(ie.whatmakesapaintingbeautiful)andperfectlyscientific(ie.wavelengthsoflight)atthesametime.Thistalkwillbeadeepdiveintothehistoryandscienceofcolorandthefascinatingworldofhumanperceptionandthescientiststhatattempttodefineit.
ManyofusarefamiliarwiththeRGBorperhapseventhesRGBcolorspace.Generally,though,westareatourcolorpickers,fiddlewiththevalues,andpicksomethingthatlooksright.However,themodernsRGBspecificationfollowsalonglineage.ItextendsallthewaybacktotheturnofthecenturyandtheCIE1931XYZcolorspace--oneoftheearliestspecifiedcolorspacesandonewhichisstillinusetoday.
Asdisplaytechnologyadvanced,sodidourcolor.Newcolorspacesemerged:CIELUV,CIELAB,YUV,HSV,HSL,RGB,sRGB.Eachwasmoldedforaspecificusecase,whetheritwasprint,perceptualuniformity,compression,orLEDdisplay.Bythetimeweworkourwaybacktothepresentday,we'releftwithalotofcoloroptionsandalotofquestions.
Inreality,though,themathandscienceisn'tallthatcomplex.We'llstartwithcolorspaceslikeRGBthatweknowandlove,andseehowthesenaturallyevolvemathematicallyfromearliercolorspacessuchasCIEXYZ.Moreimportantly,we'llseehowtoapplythisknowledgetothedecisionswemakeaboutcolorinourdaytodaywork.
JohnAustin@kleptine-gh:kleptine-AStrangerGravity
JohnAustinisadeveloperanddesignercurrentlylivinginSanFrancisco,California.Hehasbeenmakinggamesfornearly13yearsandhasworkedatGoogle,Microsoft,Funomena,andothers.Hefoundedandcurrentlyleadsthestudio,AStrangerGravity,seekingtobuildthoughtful,accessibleexperiencesthatseektoenrichthelivesofpeopleacrosstheworld.
NewprogrammingconstructsforprobabilisticAI
9/1311:20am-12:00pm,UnionStationRegencyC
Whatifitwerepossibletodostate-of-the-artAIprogrammingwithouthavingtodomath?ThistalkpresentsnovelprogrammingconstructsforprobabilisticAIthatautomateandhidedifficultmathematicalandnumericaldetailsfromtheuser.Theseconstructsareanalogoustoautomaticdifferentiationfordeeplearning,butapplytoabroaderclassofAIapproachesthatexhibitmoreaspectsof
human-likeintelligenceandcanworkwithoutanytrainingdata.
TheseconstructsalsoenableanewconceptualframeworkforprobabilisticAIthatreplacesarcanemathematicalobjectswithcodethatismoreaccessibletoprogrammerswithoutadvancedmathematicaltraining.Inthisframework,theuserfirstwritesastochasticsimulatorthatproducesimaginarydatasets.Theuserthenwritesasimplemetaprogramthatanalyzesthesimulatorcodealongsideanobservedreal-worlddataset,andinvertsthesimulatortoreturninferencesabouteventsthatcouldexplainthatdata.
ThetalkwillillustratetheseconceptsusingcodewrittenusingGen,ageneral-purposeprobabilisticprogrammingsystem.GenmakesitfeasibletousepowerfulAIapproachesthatwerepreviouslydeemedinfeasiblebecauseoftheirengineeringcomplexity.Thetalkwillshowhowtoinferthe3Dstructureandcontentofascenefromanimagebyinvertingagraphicsrendererandhowtofindthehiddencompositionalstructureintimeseriesdatatomakemoreaccurateforecasts,bothinunder100linesofcode.
MarcoCusumano-Towner-gh:marcoct-MITProbabilisticComputingProject
MarcoCusumano-Townercreatedthegeneral-purposeprobabilisticprogrammingsystemGenwhileworkingattheMITProbabilisticComputingProject.Heisafourth-yearPhDstudentinelectricalengineeringandcomputerscienceatMIT,co-advisedbyVikashMansinghkaandJoshTenenbaum.BeforeMIT,Marcoledananalyticsteamataclinicalcancerdiagnosticsstartup,andstudiedatStanfordandUCBerkeley.AsanundergraduateunderthesupervisionofPieterAbbeel,Marcodevelopedsystemsforautonomousroboticmanipulationofdeformablematerials(a.k.a.roboticlaundry-folding).Hehaspresentedhisresearchatleadingacademicconferencesinprogramminglanguages(PLDI,POPL),machinelearning(NeurIPS),androbotics(ICRA).Marcoisinterestedindevelopingprogramminglanguages,systems,userinterfaces,andothertoolsthatmakeiteasiertodevelopapplicationsofautomatedprobabilisticreasoning.
FridayLunch
9/1312:00pm-1:30pm,StifelTheatreTheater
LunchisprovidedatboththeStifelTheatreANDatUnionStation.Itisdifficulttosizethemealsappropriatelyaswearesplitbetweenthetwovenues,sowe'dpreferthatyoustayatthesamevenueasthesessionyouattendedjustpriortolunch.
StifelTheatrelunch:
Option1:RoastSirloinofBeef-Focaccia,tomatoes,sweetonionjam,BoursincheeseOption2:SmokedTurkeyandArtichokeSandwich-wholewheatbun,sundriedtomatoes,mozzarella,babyspinachleaves,basilaioliVegan(servedinKielClub):"ZLT"Sandwich-Frenchbaguette,grilledzucchini,tomatoes,sundriedtomatopestoGlutenfree(servedinKielClub):RoastsirloinofbeeforsmokedturkeySides-orangeorapple,cookie
UnionStationlunch(veg+gfoptionslabeled):
Chef'sSoupduJour(Vegan)RomaineCaesarSaladwithGarlicCrostini(deconstructed,balsamicoption)BeanSaladwithCrispVegetables&WhiteBalsamicVinaigrette(vegan)PotatoSaladtossedwithChipotleMustardDressingTheButcher'sFare,TunaSalad,GarlicSearedRoastBeefHoneyGlazedSmokedHam,RoastedTurkeyBreastGrilledVegetablesandTofuSwiss,Cheddar,Provolone&RegionalBreadPresentationGFbreadoptiononthebuffetandlabeledIndividualbagsofMs.Vickie'sPotatoChipsAssortedBrownies,Bars,&NewYorkCheesecake
Beyondtraces:theinsightsintraceaggregates
9/131:30pm-2:10pm,StifelTheatreEncore
Thevaluepropositionfordistributedtracingiswell-understood:assemblingandvisualizingend-to-endtransactionshelpstoidentifylatencybottlenecksandprovidesahead-startonproblemdiagnosis.However,traditionaltracingpracticesenabledatapresentationatthegranularityofonlyasingletransaction.Thisdataisusefulfordebuggingspecificissues,butitisdifficulttodrawconclusionsabouttheoverallsystemwithoutknowinghowrepresentativealonetraceis.Instead,aggregatingthesetracescanrevealmuchmore,andcandosowithgreaterprecisionandcertainty.
Thistalkpresentstheprofoundinsightstraceaggregateshelpunlock,includingsourcesofresourcecontention,latencyanomaliesinthecontextofserviceinfrastructure,andcorrelationsofmetricswithhigh-cardinalitycharacteristicsofthedistributedsystem.Thetalkdemonstrates,usingconcreteexamples,hownovelapplicationsofaggregatedtracesrevealnewopportunitiesforperformance
improvements.However,aggregationisnotpossiblewithoutastandardizedtracingoutputformat,aswellasaproliferationoftracesviacloud-nativeservicemeshintegration.
DanielaMiao@dymxzvf-gh:danielamiao-LightStep
DanielaMiaoiscurrentlyanEngineeringManageratLightStep,whereshejoined2yearsagoasanengineer.PriortoLightStep,shewasanengineerontheDynamoDBteamatAmazonWebServices(AWS),whereshespokeatmanyexternaleventsincludingBigDatameetupsandtheAWSdeveloperconference,re:Invent.DanielaisinterestedinvarioustopicsincludingNoSQL,privacy&securityanddistributedtracing.AtLightStep,sheiscurrentlyworkingondistributedsystemperformanceanalysis,andshespendsalotoftimethinkingabouthowtoprovidedeveloperswithvaluableperformancesignals.
TheIdeaBecomesaMachineThatMakestheArtThat...
9/131:30pm-2:10pm,StifelTheatreOvation
...MakesaMachineThatMakestheArt
SolLeWittwasaprolificAmericanartistcreditedwithfoundingConceptualArtandMinimalism.Thistalkisinspiredbyhisseriesofparticipatory,conceptualartcalled,"WallDrawings".TheWallDrawingsarelarge-scaleinstallationsthatgrowfromasetofsimplewritteninstructionsbySolLeWitthimself.
EachWallDrawingistheculminationofSolLeWittusinghisinstructionstocollaboratewithhumanassistantsthatinterprethiswritingstothebestoftheirability.LeWittoncestated,"TheIdeaBecomesaMachinethatMakestheArt."Thisbegsthequestions:Whichistheart,theidea,theexecution,orboth?Atwhatpointisthehumanthetoolandthetoolanartist?
Wewillcontinuethismodeofthoughtandquestionwhethertheworkgeneratedbyamachinetrainedbyhumans(followingtheinstructionsoftheartist,SolLeWitt)canbeconsideredart.Wewillexploreahuman-machinecollaborationusinghumanandAIassistantsthatwillinstallinterpretationsofWallDrawingsinavirtualspaceonline.
WewillgivedifferentWallDrawingpromptstohumanassistantssotheycaninterpretandcreateversionsofSolLeWitt'sart.Giventhedatafromthesehumanassistants,wewilltrainneuralnetworkstogeneratetheirownworkbasedonthesepromptsusingmachinelearningtechniquessuchasRNN,LSTM,andGAN.Finally,weplantofeedthesebackintoacollaborativeenvironmentforbothhumanandAIassistants.
ChristineStavridis-gh:promptandpleasant-Prompt&Pleasant
Prompt&PleasantisadigitalpracticecomprisedofChristineStavridisandJonathanKing.Weareexploringwaystomaketheseeminglyinaccessible,accessible.
JonathanKing-gh:promptandpleasant-Prompt&Pleasant
Prompt&PleasantisadigitalpracticecomprisedofChristineStavridisandJonathanKing.Weareexploringwaystomaketheseeminglyinaccessible,accessible.
ApachePulsarandtheLegacyofTelegraphy
9/131:30pm-2:10pm,StifelTheatrePeabodyLounge
TheWirelessTelegraphwasinvitedinthe1890sandquicklyspreadtobethedominantmethodofpeer-to-peercommunicationthroughoutthe19thcentury.Asamethodofcommunication,theTelegraphallowedforasecure,Intercontinental,high-throughputandextensiblemessagetransmission.Towardtheendofitspopularity,Telegraphswereabroadcastprotocol(onesendertomanyrecipients),andtheirutilityskyrocketed.ThedesignconsiderationsandengineeringoftheTelegraphareseenincontemporarymessagingsystems,especiallyApachePulsar.Thistalkcoversthedesignconsiderationsrequiredtoenabledgloballydistributed,low-latencyapplicationstorunontopofApachePulsar.Fromthestoragemodeltoreplicationanddeployment,Iexploreeachofthesedesigndecisionsandhowtheycontrastwiththedesignofwirelesstelegraphy.ThetalkconcludeswithademoofApachePulsarfunctionsandPulsarSQL,displayingtheextensibilityofPulsar.AttendeescanexpecttolearnabriefhistoryofwirelesstelegraphyandadeepunderstandingofthedesignofApachePulsar.
JowanzaJoseph@Jowanza-gh:josep2-Pluralsight
JowanzaisasoftwareengineeratPluralsight.PrevioustoPluralsight,JowanzaworkedatAdobe,Zagg,andEdgebyAscentialonmachinelearning,distributedmicroservicesanddistributeddataprocessing.JowanaisafrequentconferencespeakerincludingappearancesatOpenSourceSummit,O'ReillyStrata,ScalabytheBay,OpenWestandAllThingsOpen.Jowanzacontributestoseveralopensourceprojectsandisanactivebloggeronhiswebsite.
VoiceDrivenDevelopment:Whoneedsakeyboardanyway?
9/131:30pm-2:10pm,StifelTheatrePremier
Asaprogrammer,beingunabletouseakeyboardcanseemlikeacareer-endinglimitation.IwasfacingthatrealityayearagoduetoRepetitiveStrainInjury,wheneveryproductivedaywasastepbackformyhealth.Inafutile,last-ditcheffortatsavingmycareer,Iturnedtospeechrecognition.Unexpectedly,notonlywasIabletoreturntopain-freeproductivity,butIfoundvoice-drivendevelopmenttobeenjoyable,efficient,andasourceofrenewedexcitementforprogramming.Speechrecognitionisn'texactlyknowninsoftwaredevelopmentcirclesasaworkableapproachtoprogramming,butit'sbetterthanyou(probably)think!Farfromimposingatediousworkflowwithaspecializedsetofcommands,speechcanenableflexibilityandoptimizationsthataredifficulttoachievewithakeyboard,anddoesnotgetinthewaywhenadaptingtonewtechnologies,tools,orlanguages.Thispractical,demo-driventalkpresentstheapproachandtoolsthathaveallowedmetocontinueonwithmycareer.I'llpresentrealworldexperiencewithspeech-relatedtopicslikedisambiguatinghomophonesontheflyandteachingthevoiceenginenewtechnicalvocabulary.Beyondtechnicalaspects,Iwillsharemyexperienceguidingateamtosupportthisworkingstyle,andchallengestoaddingamicrophoneintoanopenofficeenvironment.Speechnotonlyremovesthecompromisebetweenhealthandcareerwhenaninjuryispresent,butcanalsoaugmentandimprovetheworkflowforawideraudience.
EmilyShea@yomilly-gh:2shea-Fastly,Inc.
EmilyisaSeniorSoftwareEngineeratFastly,wheresheworksontheplatformfordeliveringcoreCDNconfigurations,anddevelopsinPerlusingspeechrecognition.Inapastlife,sheworkedinHRatmobilegamingcompanies.EmilyholdsaBAinArchitecturefromUCBerkeley,andinhersparetimelikestohangoutinparkswithherdog,namedChicken.
Findingbugswithoutrunningorevenlookingatcode
9/131:30pm-2:10pm,StifelTheatreTheater
Whatifyoucouldfindcomplexbugsinsystemswithouteverhavinglookedatanyofthecode,withoutrunningthecode,withoutcloningthecode,orevenknowingwhatlanguagethecodeiswritteninorwhereitsgitrepolives?Whatifyoucouldvalidatethecorrectnessofanarchitecturalproposalbeforewritingcode?
Whilethismightsoundimpossible,it'sinfactquitedoable,andhappeningeveryday!
Toolscalled"modelcheckers"canbeusedtomodelandanalyzethearchitecture/designofasystem.Modelcheckerscansimulateeverypossiblesequenceofoperationsyoursystemmighteverencounter,andcheckifdesiredpropertiesalwayshold.Andifapropertydoesn'thold,themodelcheckerwillshowyouanexactsequencesofoperationsthatviolatesit.
Benefitcanbehadbothintheearlydesignstagesofaproject,orifyou'reanewcomertryingtogetup-to-speedonanexistingsystem.Examplesofbothwillbeshown.
ThistalkwilldescribethemodelcheckerAlloy(http://alloytools.org).I'llshowhowweusedittoinvalidatealargeamountofworkbyonegroup(beforemultipleteamsstartedtotryintegratingwiththatwork).I'llalsoshowhowweemployedittofindahighlysubtlesecurityflawinanothergroup'sproject,withouteverlookingatasinglelineofcodefromeithergroup.Inbothcases,wedidthissimplybytalkingtoteamsaboutthearchitectureoftheirsystems,andtranslatingthosedesignsintoformalspecifications.
JayParlar@parlar-gh:parlarjb-Okta
JayParlarhasbeenasoftwaredeveloperfor~20years,andholdsaPhDinSoftwareEngineering.Hismajorfocusthepast~3yearshasbeenontheÂapplicationofformalmethodstorealindustryproblems.ThishasresultedininterestingsuccesseswithAlloy,TLA+,andinwritingacustommodelcheckerinZ3.Buthe'dalsolovetotalktoyouaboutPython,runningandfountainpens!
Hownottoreadtheroom:CreatingwearableswithML
9/131:30pm-2:10pm,UnionStationGrandF
I'manintrovert.Thiscanbeabitunfortunate,whenyouareapersonthatenjoysspendingalotoffreetimecreatingfashionthingsbedazzledwithLEDs...onlytorarelywearthemoutinpublic.ÂInanefforttoactuallysharemyweirdandwonderfulcreationswithothers,Idecidedtocreateawearableprojectthatwouldforcemetobesociableinorderforittorevealitsmagic.Inthistalk,I'llsharehowIamusingmachinelearningwithjavascriptandtinycomputerstomake"fashion"thatisresponsivetothepeoplearoundyouandtheattentionyouare(oraren't)receiving.
StephanieNemeth@stephaniecodes-gh:traumverloren-Microsoft
StephanieisadeveloperlivinginBerlin.SheenjoysexperimentingwithhardwareandLEDstomakebeautiful,uselessthings.
AStitchinTime-TheFutureofOSSSustainability
9/131:30pm-2:10pm,UnionStationRegencyAB
OpenSourceSoftware(OSS)hascreatedenormousvalueforindividualsandbusinesses.Howeverthecreatorsofthatsoftwareoftencaptureatinyfractionofthatvalue,andcanenduptired,burntout,orquittingtheprojectentirely.Inthepastfewyears,newfearshavearisenthatcloudproviderswill'take'opensourcesoftwareandresellit,leadingsomecompaniestomoveawayfromopen-sourcelicensestopreventthisfromhappening.
ClojuristsTogetherisanorganisationcreatedtohelpsupportandgrowcriticalopensourceClojuresoftware.ThistalkwilldiscussourexperiencegrowingClojuristsTogetherandsomeinsightsithasgivenusinto:
ThebenefitsoffundingOSSforallpartiesDifferentwaysoffundingOSS,andhowtheysuitdifferentkindsofprojectsHowdifferentlicensesimpactfundingmodels,andwhyyoushouldpickalicensecarefully!WhatbusinessesarelookingforwhentheysponsoropensourceLessonswe'velearntservingtheClojurecommunityFuturedirectionsforfundingOSSthatwethinkarepromising
DanielCompton@danielwithmusic-gh:danielcompton-Falcon
DanielComptonisasoftwareengineerlivinginMorrinsvilleNewZealand.HehasbeeninvolvedintheClojurecommunitysince2014,workingonanumberofopensourceprojectsincludingKibit,re-frame,andre-frame-10x.HeistheprojectleaderofClojuristsTogether,acommunityorganisationdedicatedtofundingcriticalClojureopensourcesoftware.HealsopublishesTheREPL,anewsletterandpodcastaboutClojure.Heisengineer#1atTheFalconProject,astealthstartup.PriortothatheworkedatDay8buildingsystemsinClojureandClojureScript.
PrivacyGovernance&ExplainabilityinML/AI
9/131:30pm-2:10pm,UnionStationRegencyC
SincetheGeneralDataProtectionRegulation(GDPR)wentintoeffectinMay2018,mattersofdataprivacyhavegrownfromminororganizationaladjustmentstoenterprise-levelinitiativeswithimpactoninnovationandday-to-dayoperationsalike.Whileprivacycompliancemaybestraightforwardinsomeareas,thegrowthandexpansionofmachinelearning(ML)andartificialintelligence(AI)havehascreatedanimpassebetweenconsumerdataandprocessesthatare,tosaytheleast,difficulttofullyexplain.IntegratingprocesseswithMLandAItechniquesoftenprovetosignificantlybenefittheaccuracyandefficiencyofprocessesanddecisionmaking,butone'sabilitytofullyunderstandpreciselyhowanoutputwasgeneratedoradecisionwasmadeforanindividualismucheasiersaidthandone.Yet,regulatorsacrosstheglobearechallengingbusinessestoexplainhowtheyareusinggovernancetechniquestoprotectconsumerdataprivacyandtoexplainhowdecisionmakingwithinML/AIisimpactingconsumers.Howcanoneidentifybias?Whatprocessescanbeintroducedtoprotectconsumerprivacywhilerootingoutpotentialbiasintheunderlyingmodels?Inthistalk,wewillexploremethodsforenhancingprivacyandgoverningdatathatisusedforML/AI,aswellastoconsiderproceduralapproachesavailableforrootingoutbiasandbuildingafoundationforconsumerconfidenceinanotherwisecomplexandopaquespace.
JaredMaslin-SlalomConsulting
JaredisaSolutionArchitectwithSlalomConsultinginSt.Louis,aswellasanEducatorwiththeUniversityofCalifornia,Berkeley,wherehesupportsacourseonHumanValuesandEthicsinDataScience.JaredhasmorethanadecadeofdiverseexperienceinDataPrivacy,DataAnalytics,Auditing,Compliance,andFinance,whichhascontributedtoauniqueperspectiveinthespace.
DeterministicDockerImageswithGoMicroservices
9/132:30pm-3:10pm,StifelTheatreEncore
AtSamsara,wecontinuouslydeploymorethanahundreddifferentmicroservices,allofwhichwewantedtodoaseasilyandautomaticallyaspossible.Wewantedtomakethisevenbetterbybringingourdeploytimestounder20minuteswhileminimizingdisruption,asthiswouldalsomeanthatwewouldstarttodeploynearlyourentirestackabout3timesahour.Oursolutiontothat:deterministicallybuildingourservicesinourdeployedimagestominimizetimespentbuildingservices,networkcostswithmovingimagesaround,butmostimportantly,beingabletoautomaticallyfigureoutwhichofourserviceshaveactuallychangedandonlydeployingthose.
Thisofcourse,comeswithsomechallengesfromtheinfrastructurelayer,tothecodepatternsanddesignweuseacrosstheorganization.
We'llbetalkingaboutthesechallengesduringthistalk,namely:
1. Buildingdeterministicgolangbinaries,andtheedgecasesandflagsyouhavetodotogetitright,aswellassomehistoryongolangdevelopment'seffortstobedeterministicandtimeswhenit'sregressed.
2. Buildingdeterministicdockerimageswithbazel,andadiveintohowdockerworksandwhyitdoesn'tnativelysupportthis.3. Challengeswithkeepingcodedecoupledtopreventchangesfromrebuildingallofourimages,aswellasthepatternsandtoolswe
usetohelpwiththis.
JonSanMiguel-gh:jsm-Samsara
SeniorSoftwareEngineer@Samsara,previouslyatHuddleUp,Optimizely,Yahoo,Lexity,&Servio.Beenthroughmanylifespansofstartups,seeingthembothfail&growtremendously,withextensiveexperiencerangingfrombuildingimmersivefrontends,torobustlow-levelinfrastructure.
ImprovinglawinterpretabilityusingNLP
9/132:30pm-3:10pm,StifelTheatreOvation
Theprocessoflegalreasoningisheavilyreliantoninformationstoredintext,butwhilelegaltextsaregenerallyeasilyaccessible,theirinterpretationoftenisn'tstraightforward,makingtheunderstandingofthelaweffectivelyinaccessibletothegeneralpublic.
DataScientistsfromBardess,incollaborationwitharesearchgroupfromtheGovernmentofOntario,haveinvestigatedhowNaturalLanguageProcessingtechniquescanbeappliedtounderstandlinguisticpatternsinlegislativetextsandextractinformationthatismeaningfulforthepublic.
UsingtheAccessibilityforOntarianswithDisabilitiesAct(AODA)asatestcase,wedevelopedamulti-stageanalysisthatcombinessomewellknownNLPmethodologiesinauniqueapproach.UltimatelywewereabletoautomatetheextractionofrulesfromtheActanditsRegulation,identifytheentitiesresponsibleforcompliance,andorganizethemintogroupsthatarehomogeneouswithrespecttotheirimpactonvariousentitiesandindustries.
Themethodologydevelopedprovidesuswithaframeworkforrepresentinglegaltextsthatcanbeusedtosimplifythewayinformationinthelawisaccessedbythepublicandatthesametimehighlightspartsofthelawthatareparticularlyhardtointerpretandshouldbere-writtenmoreclearly.
SerenaPeruzzo@sereprz-gh:sereprz-BardessGroup
SerenaisaseniordatascientistattheanalyticsconsultancyBardess,currentlybasedinToronto,Canada.BeforejoiningBardess,shehasworkedbothinacademiaasanMLresearcherandintheindustryasadatascienceconsultantontheAustralian,BritishandCanadianmarkets.Serenaispassionateabouteducation,communityandtechforgoodandshesplitsherfreetimebetweenmentoringdatasciencestudents,organizingmeetupsandvolunteering.
Alda'sdynamicrelationshipwithClojure
9/132:30pm-3:10pm,StifelTheatrePeabodyLounge
Aldaisamusicprogramminglanguagethatprovidesawaytodescribeamusicalcompositionasatextdocument.Itssyntaxiseasytouseandwell-suitedforbothmusicianswithlittle-to-noprogrammingknowledgeandprogrammerswholackmusictheoryknowledge.
AldaandClojurehaveacomplexrelationshipthatisatthecenterofthistalk.Iwilltellthestoryofhowthetwobecameintertwinedandmycurrentefforttodisentanglethem.InterestingtopicsthatwillbediscussedalongthewayincludeZeroMQ,OpenSoundControl(OSC),andthecreationofnativeexecutables.
DaveYarwood@dave_yarwood-gh:daveyarwood-Adzerk
DaveYarwoodisacomposer,musician,andsoftwareengineeratAdzerk,whereheusesClojureandClojureScripttobuildUIs,APIs,andbackendservices.InadditiontoAlda(anopen-sourcemusiccompositionprogramminglanguage),hemaintainsanumberofopen-sourceClojurelibrariessuchasezzmq(ahigh-levelZeroMQlibraryforClojure)andMantra(aClojureScriptWebAudioAPImusiclibrary).
DaveholdsaBachelorofMusicdegreefromtheUniversityofNorthCarolinaatChapelHill,withfocusesinMusicCompositionandBassoonPerformance.Heisendlesslyfascinatedbytheintersectingworldsofmusicandprogramming.
DigitalandSocialResiliencethroughtheNYCMesh
9/132:30pm-3:10pm,StifelTheatrePremier
NYCMeshisacommunity-runmeshnetworkthathelpsoverthree-hundredhomesconnecttohigh-speedInternet;makingitoneofthelargestmeshnetworksintheworld.WeworkwithanyoneinterestedinreclaimingownershipovertheirInternet,butspecificallyengagewithunderservedpopulationsincludingresidentsinlow-incomehousingandcommunitycenters.Builtbyacombinationofconsumer-gradematerialsanddonatedfiber-opticcable,thisnetworkgivespeopleownershipandcontrolovertheirInternetusage.
Digitalempowermentdoesnotendattheconnection.AtNYCMesh,weteachmembershowtobuildandmaintainthenetwork
themselves,resultinginatrulydecentralizedandresilientnetwork.Ourtrainingprogramteachesmemberstoleadnodeinstallations,ataskthatrequiresplanning,criticalthought,andcooperationfromtheteam,givingourmembersandeeperappreciationforthisutility.
Inthistalk,wewilldescribehowwehaveappliedthenotionofdecentralizationboththetechnicalandorganizationaldesignoftheNYCMesh.WewilltalkaboutwhyeducatingourcommunityonthenetworkinglayerisessentialinordertobuildatrulyequitableInternetandhowourtrainingprogramhelpsdevelopinformedusers.Wewillalsodescribeourgoalsforexpandingthenetworkanditspotentialforcommunitybuildingwithhyper-localizedwebsitesandapplications.
MichaelDonatz@donutsonhudson-gh:donutsonhudson-BloombergLP
MichaelisavolunteerwithNYCMeshandtheOutinTechDigitalCorps.BydayheworksinPhilanthropyEngineeringatBloomberg.
JillianMurphy-NYCMesh
JillianMurphyvolunteerswithNYCMeshwithaparticularpassionforbreakingdownsocialbarrierswiththeaidsoftechnologyandeducation.Jillianisaninstallleader,eventorganizer,andauthoroftheorganization'smonthlynewsletter.TheIRSknowsherasanadministratoratNewYorkUniversitywheresheisalsostudyingbusiness.
DeclarativeAPIsinanImperativeWorld
9/132:30pm-3:10pm,StifelTheatreTheater
DeclarativeUIframeworkshavetakenovertheJavaScriptlandscape.Whichisgreat,becausemakingassumptionsaboutyourapplication'sstateisasourceofpain,frustration,andbugs.ButwhathappenswhenyouneedtogluetogetherimperativeAPIswithyourdeclarativeUIframework?
Inthistalk,I'llexplainproblemstheAtomEditorteamcameacrosswhenwestartedusingReactasourUIframework,aswellasthesolutionswecameupwith.We'lldiscusshowtogeneralizethesepatternstointegratewithotherimperativeAPIsyoumightcomeacrossinthewild,suchasanimationlibraries.Finally,we'llexplorewhatwecanlearnfromthehistoryofsoftwaretrendsovertimetocreateabetterfuturetogether.
TildeThurium@annthurium-gh:annthurium
TildeAnnThuriumisanartist,engiqueer,andactivist.Inpreviouslives,theyhavedonestintsasaflorist,asecurityguard,andaHumanResourceswench.Intheirsparetime,theyillustratedatastructuresandalgorithmswithacrylicpaint.
Howtobuildwithdatatrustandprivacyasabaseline
9/132:30pm-3:10pm,UnionStationGrandF
TimBerners-LeerecentlypublishedhisContractfortheWebwithacoreprinciplestatingwemust"Respectconsumers'privacyandpersonaldatasopeopleareincontroloftheirlivesonline."
Noble'stalkexplorestacticalapproachestobegoodstewardsofdatainthefaceofgrowingregulationsthroughsoftwarewithtechniqueslikedifferentialprivacy.
NobleAckerson@nobleackerson-gh:stigsfoot-OpenARCloud
NobleisaSeniorManagerforProductManagementandSoftwareatOpenARCloudwhereheadvisesonDataGovernance,DataPrivacy,DataTrusttoteamsacross55countries.He'salsoaGoogleDevelopersExpertforproductstrategyandarecoveringstartupfounderdeliveringaugmentedreality,virtualreality,webandmobiletechnologiesfromabstractconceptstofinalproducts.
NoblesitsontheAlumniLeadershipBoardfortheleadingonlineeducationstartup,Udacity.AmemberoftheAlumniBoardathisalmamater,LynchburgCollegeandactsasthePublicityChairforUbicompandISWC,twotop-tierconferencesfortheinternetofthingsandwearabletechnology.WhenNobleisnotfocusedonproduct,heismentoringunderrepresentedandaspiringdevelopers,speakingattechnologyeventslikethisone,orspendingtimewithhisdaughters.
AutomatingOperationsofaGlobalCDN
9/132:30pm-3:10pm,UnionStationRegencyAB
AspectsoftheNetflixgrowthstoryareveryvisible-thetransitionfromphysicalmediadeliverytodigitalstreaming,orfromadomesticfocustoaglobalone.However,therearesometransitionsthatmaybelessnoticeable.
OneexampleoftheseunnoticedtransitionsisfromrelianceoncommercialCDNprovidersforcontentdelivery,toutilizinganin-house,purpose-builtglobalCDNknownasOpenConnect.ThisisthesoleproviderofallNetflixstreamingcontent.
EfficientlyoperatingthisglobalCDNrequiresmorethanjustcompetentpersonnel.Thereisheavyrelianceonsoftwareautomationandtooling,whichhelpstheCDNscaleatafasterpacethantheheadcountrequiredtorunit.
Giventhefast-pacedgrowthofNetflixstreaming,therehavebeenmanychallengesinhowtobestsupportOpenConnectoperationswithautomation.Teamsemergedtocollectivelybuildapplicationsthathelpmeettheseneeds,butovertimetheseapplicationsballoonedtorelativelylarge,siloedmonolithsthatarefunctional,butoftendifficulttodeployandmanage.
Throughthistalk,youwillreceiveinsightintohowwehavestreamlinedourautomationstory.Howweactivelytransitionedfromlargemonolithstoamoredecomposed,scalablesetofmicroserviceswerefertoastheOpenConnectAdministrationPlatform.Youwillbewalkedthroughourautomationjourney-ourhistory,ourfuture,thepitfallswehaveencounteredandthelessonswehavelearnedalongtheway.
RobertFernandes@bovilexic-Netflix
Forthelast2yearsRobertFernandeshasleadtheOpenConnectToolsteamwithinNetflix,whichhandlesthesoftwareautomationandtoolingfortheNetflixglobalCDN.Priortomanagingthisteam,hewasacontributingengineerontheteamforover4years-helpingtobuildandmaintainalargeportionofthesuiteofsolutionsittakestooperatetheCDN.PriortoNetflix,RobertspentoveradecadeinthegovernmentR&Dspaceworkingforanationalresearchlaboratory,followedbyafewyearsintheareaofenterprisebusinesscollaborationsoftware.Hisentirecareerhasbeenspentasanarchitectandageneralist,workingonallpartsofthestack-fromdataandpersistence,tomiddletierservicesandAPIs,allthewayuptofrontenddesignanddevelopmentusingavarietyofUIframeworks.
Ridingthestreamprocessingwave
9/132:30pm-3:10pm,UnionStationRegencyC
AtLinkedIn,werunseveralthousandsofstreamprocessingapplicationswhich,coupledwithourscale,hasexposedustosomeuniquechallenges.Wewilltalkaboutthe3kindsofapplicationsthathavemadethemostimpactonourstreamprocessingplatform.
MachineLearningapplicationsaredrivingsomeofthelatestinnovationsforstreaming.Thecurrenttrendistotrainamodelinbatchenvironmentsanddoinferenceinonlineenvironments.Webuiltsomenativecapabilitiessuchas"side-inputs"forhandlinglargestate,whileallowingfeaturestobecontinuouslypushedfromofflinegridstostreamingenvironments.DataScientistspreferDSL'sforfeaturegenerationandaccess.Consequently,webuilttheabilitytoconvertamachinelearningDSLtoastreamingjobanduseitforfeatureengineering.WewilltalkaboutthiscapabilityandhowthiscanbeextendedtoconvertHive,PigorothercustomDSLtostreamingapplications.
Wehaveobservedtheemergenceofapplicationsthataremovingfrombatchprocessingmodetonearlineprocessingmodeaswellasoperatingonbothbatch(HDFS)andstreaming(Kafka)datasets(e.gExperimentation).AtLinkedIn,weuseSamzaforstreamprocessing,andSamzaapplicationscanachieveoffline-onlineconvergenceofstreamandbatchprocessingbysimplyswitchingthestreaminginputsystemslikeKafkawithHDFS-basedinput.ApacheBeamintegrationforSamzaenablesthecapabilitytoexecuteindifferentenvironments.Streamingapplicationsnowmaintainverylargelocalstate,andduringdeployments,applicationornodefailuresitiscriticaltorestorethisstatetoitspreviousversion.Wewilltalkabouttheimpactofthesefailuresonlargestatefulapplicationsandsomeoftherecentimprovementswehavemadeinhostaffinity,staterestoreandournewstandbycontainersolution.
SamarthShetty-LinkedIn
SamarthShettyisanengineeringleaderwith14+yearsofexperienceinbuildingglobalscalecloudinfrastructure.AtLinkedIn,heleadstheStreamprocessing(ApacheSamza)andDatapipeline(BrooklinandDatabus)teams.PriortoLinkedIn,heleddevelopmentteamsthatbuiltcorecomponentsforAzureStorage,AzureCDN,andWindowsatMicrosoft.
SafetyinChaos:FormingRealisticFailureHypotheses
9/133:30pm-4:10pm,StifelTheatreEncore
Subjectingsystemstofailuresissupposedtoincreaseconfidenceintheirstability.Butwhy?Howdoyouformfailurehypotheses?Howdoyoureasonabouttheirsafety?Whyshouldyourorganizationlistentoyouandinvestintestingyourfailurehypotheses?
ThesearesomeofthequestionsIfacedduringmyquesttoimproveproductionstabilityatwork.Inthistalk,wewilldiscussthreequestions:
Howtoformbetterhypotheses,andinparticular,howtoseparateopinionsfromhypotheses?Iwillsharesomeexamplesofoverlysimplistichypotheses,testingofwhichmaynotproducedesiredoutcomes.Howtopushsafetyboundariestohelpyouincreasetheriskinessofyourhypotheses?Iwilloutlinesometenetsofincreasingsystemssafety,beforeyoucanincreasetheriskinessofyourhypothesis.Howtoinfluencebettertrade-offsbetweeninvestinginchaosengineeringandeverythingelse?Howtoreleasetheconstanttrade-offtensionthatexistsinmostorganizations,sothat,youasengineer,caninfluenceyourorganizationtoinvestinchaostesting?
SubbuAllamaraju@sallamar-gh:S3u-ExpediaGroup
SubbuAllamarajuisaseniortechnologistattheExpediaGroup,whereheisleadingalarge-scalemigrationofExpediaGroup'stravelplatformsfromenterprisedatacenterstoahighlyavailablearchitectureinthecloud.Subbuisawell-roundedengineerandinfluencerwithhands-onexperienceinsoftwaredevelopment,architecture,distributedsystems,services,internetprotocols,operations,andthecloud.Overthepastseveralyears,hehashelpedbuildandempowerseveralengineeringandoperationsteamsintheseareas.
TowardsanOpen,Reason(ML)ableWeb
9/133:30pm-4:10pm,StifelTheatreOvation
Webtechnologiestodayareconvergingtowardstwoparadigms-statictypingandcompilationtoJavaScript.Morethaneverbefore,engineerswantsoundtypesystemsandelegantlanguageergonomicswhilegeneratingperformantJS.ReasonMLpromisesthebestofbothbypairingOCaml'srobusttypesystemwithablazingfastJavaScriptcompiler,BuckleScript.
Despitethesebenefits,Reasonmadeaminimalsplashwhenitarrivedin2016.OtherprojectslikeTypeScripteclipsedcommunityattentionaroundstatictypingandperformance.Butthat'sbeginningtochange.Accordingto2018'sStateofJSsurvey,Reasonhasseen94%growthindeveloperinterestoverthelastyear,greaterthananyothersurveyedlanguage.Thissuccesshasbeentheresultofadistributedopensourceeffortbycommunitycontributors.TheproliferationofbindingsforJavaScriptlibrarieslikeReactandApolloaremakingReasonaproductionchoiceforteams.Inaddition,thecreationofanativepackagemanager,esy,andUIframework,revery,signalsamajorbreakthroughincross-platformdevelopment.
Inthistalk,we'lldiveintohowtheReasoncommunityembracedopensourcedesigntocreatealanguagethatfulfillsthepromiseof"writeonce,runanywhere."We'lldigintoReason'sinternalsandexaminethemanyrapidevolutionsofthelanguage.Finally,we'lldiscussthechallengesofdevelopingopensourceinunchartedtechnicalfrontiersthroughthelensoftwoReasonprojects,reveryandwonka.
ParkerZiegler@parker_ziegler-gh:parkerziegler-Formidable
ParkerisasoftwareengineeratFormidable,aJavaScriptconsultancyandopensourcesoftwareorganization.HehelpstoauthorandmaintainFormidable'sexpansiveopensourcerepertoire,andcontributesactivelytoprojectslikevictory,spectacle,webpack-dashboard,andurql.Overthepastyear,ParkerhasbecomeanactivecontributorintheReasoncommunityworkingonreason-react,revery,andwonka.Mostrecentlyhehasauthoredreason-urql,afully-featuredGraphQLclientforReasonandOCaml.
EnhancingAngklungMusicRehearsalswithModernTech
9/133:30pm-4:10pm,StifelTheatrePeabodyLounge
AngklungisatraditionalmusicalinstrumentfromIndonesia.Thisinstrumenthasalotofvarietyinhowitisperformed;acommonformatistheorchestralformatinwhich15-30playersgathertoformateam.Playingangklunginthiswayisfunbutalsopresentssomechallengesthatarehardtosolvemanually.Inthistalk,wewilllearnhowtechnologyisusedtoimprovethequalityofrehearsalsofthisage-oldinstrument.
First,IwillintroducetheAngklung,andexplainthekeyconceptsofanAngklungteam.Then,IwillshowyouthemanualprocessesinvolvedinforminganAngklungteam,andhowthoseprocessescanbeautomated,increasingefficiencyofrehearsalsandqualityofthemusicitself,withthehelpoflanguage-agnosticalgorithmsinterfacedbyawebapplication.ThisinvolvesreadingAngklung-specificsheetmusicofasong(writtenincipher/numbernotation),assessingthedistributionofnotesinthatsong,andfinallycalculatingthemostoptimaldistributionofAngklungforateambasedonseveralimportantfactors.IwillalsoshowsomeactualcodesamplesanddemonstrateanapplicationofthealgorithmwritteninPython,toshowhowanAngklungteamisformedinreallife.
TrapsiloBumi-gh:tbumi-HENNGE
BumiisaSoftwareDeveloperforHENNGE,basedinTokyo,Japan.He'salwaysexcitedinchallengeswhetheritbetinkeringwithnewtechorsharingitwithothersthroughtalks.Whenhe'snotcoding,hecanbeseentravelling,boardgaming,orplaying(andconducting)Angklung.
AssistiveAugmentation:LipReadingwithAI
9/133:30pm-4:10pm,StifelTheatrePremier
IntheUSalone,approximately3%ofthepopulation(10million)areeitherdeaforhavemoderatetoprofoundhearingloss.Thisis3timesasmanypeoplethanthoseinwheelchairsyetreasonabledisabilityaccommodationsforthedeaforhearingimpairedonlyrequireanASL(AmericanSignLanguage)interpreterincertaincircumstancessuchinofficialpolitical,legal,education,lawenforcement,andemploymenteventsandsituations.Theproblemwiththisisthatonlyafractionofthefunctionallydeaf(250-500thousand)speakASL(alsocalled"signers")andsituationsthathaveaccommodationsarefewandfarin-betweenanywaysohowcanthehearingimpaired
engagewitheventsliketherestofusandespeciallythosethatdon'thaveinterpreterssuchasmeetups,conferences,anddebates.
Existingsolutionsfocusstrictlyonprovidingclosedcaptionservices,whenavailable,videorelayservices,whicharemoresuitablefortwowaycommunication,orspeechrecognitionwhichcanworkverywellbutsincetheyappearonadifferentscreendon'tallowthedeafpersontoengagewiththeeventasjustanotheraudiencemember.
Howcandeeplearningmethodssolvethisproblemwithlipreading?
SergMasis-gh:smasis001-IllinoisInstituteofTechnology
DataScientistat3DprintermanufacturerFormlabs,agraduatestudentatIllinoisTech,acomputerscientist,aformerwebmasterforoneofthelargestpokersitesintheworld,thefounderofasearchenginestartupfocusedonuser-behavioranddecision-makingandsomeonepassionateaboutusingdataandtechnologytoimprovepeopleslives.
Uptime15,364days-TheComputersofVoyager
9/133:30pm-4:10pm,StifelTheatreTheater
TheVoyager1andVoyager2spaceprobes,bothlaunchedin1977,eachhadaprimaryobjectivetoexploreJupiterandSaturn.Thisgoalwasachievedby1981.YetVoyager,NASA'slongestrunningmission,hascontinuedtothisday.BothVoyagerprobesarestilloperating,andreturningscientificdatafromoutsideoursolarsystem.
ThistalkexploresthecomputingsystemsofVoyager-thesystemswhichenableremotecontrolofthespacecraft,andprovidefortherecordingandreturnofdatatoEarth.Thesesystemshaveprovedtobebothadaptable,durable,andresilientinsupportofascientificundertakingnowinit'sfifthdecade.
WhatcanwelearnfromtheengineeringofVoyager'scomputingsystems?Whyhavetheysurvivedforsolongintheharshenvironmentofspace?Whatisinvolvedinpatchingasystemfromabillionmilesaway?Andwhatdoesthefuturehold?
AaronCummings@btvaaron-gh:aaroncummings
AaronCummingsisasoftwaredeveloperworkinginthesemiconductorindustry,currentlyleadingateamworkingontoolsforbuildingandtestingembeddedmemories.Hehashadalongtermfascinationwiththespaceprogram,andhasbeeninterestedinVoyagersinceseeingthepicturesreturnedfromJupiterandSaturninthe1980s.
FormalSpecificationandTamingOtherPeople'sTech
9/133:30pm-4:10pm,UnionStationGrandF
FormalspecificationsarenotjustforacademicsandhaveapplicationswaybeyondcheckingalgorithmslikePaxosandRaft.Joinusforatourofusingformalspecificationsforeverydaysystemdesign.Ifyou'veeverconfiguredaloadbalanceroramessagequeuethenyou'realreadydoingthekindofengineeringworkthatformalspecificationswereintendedfor.Thistalktellsthestoryofareallifeinproductionsystembuiltwithcommonopensourcetechnologiesthatwasstrugglingwithgrowthandhowweuntangleditsbuilt-inassumptionsandreworkeditsdesignusingformalspecifications.
MarianneBellotti@bellmar-gh:mbellotti-Auth0
MarianneBellottirunstheInternalServicesteamatAuth0,whichworkscloselywithSREtobuildhighlyscalableservicesforotherengineeringteams.Shehasanobsessiveloveofcomplexstrugglingsystems--theolderthebetter--andspentthreeyearsrescuingsuchsystemsfortheFederalgovernmentaspartofUnitedStatesDigitalService.
Thestrangeloopmakingparasiticinsectsourfriends
9/133:30pm-4:10pm,UnionStationRegencyAB
Thehistoryofecologyandfunctionalprogrammingareintertwined.TheFibonaccisequence,commonlyusedtoillustraterecursionforlearnersoffunctionalprogramming,wasinitiallycreatedtomodelapopulationofrabbits.In1935,theentomologistJohnNicholsonandthephysicistVictorBaileypublishedTheBalanceofAnimalPopulations.Theyproposedequationstorepresentthepopulationinteractionbetweenaparasiticwaspanditsinsecthost.AttheheartoftheNicholson-Baileymodelisamutuallyrecursiverelationship:astrangeloop.Thetwopopulationsdependuponeachotherandwillrecurseforeverorcrashastheybothapproachzero.Themodelbecameaparadigminecologicalmodelingandisanessentialtoolinthefieldofbiologicalcontrol,whereinsectparasitesthatlaytheireggsinsideaninvasivepestspeciesareaugmentedforpestmanagement.However,Nicholson-Baileytypemodelsusedinbiologicalcontrolstudiesrarelyutilizehigh-levelprogramminglanguagetools.Generationsisanopen-sourcePythonpackagecontainingparameterizablemodulesforunderstandingthepopulationdynamicsofbiologicalcontrolinsectsandtheirtargetplants(invasiveweeds).Generationsmodelstheresponseofaninvasiveplantpopulationtoanherbivorousparasite,throughasetofrecursivefunctions.Comelearnhowfunctionalprogrammingisusedtofindorganicsolutionstoinvasivespeciesproblemstosavenativeecosystemsandminimize
theuseofpesticides.
MaryMarek-Spartz@mMarekSpartz-gh:alfalimajuliett-UniversityofMinnesota,TwinCities
IamanEntomologyPhDstudentattheUniversityofMinnesotaworkingonweedbiologicalcontrolprojects.IhaveaMaster'sdegreefromSt.Mary'sUniversityofMinnesotainGeographicInformationScienceandIhavebeenanadjunctinstructortheresince2016,teachingprogrammingprinciplesforGISanddatainformaticsandgeospatialanalysis.IhaverecentlybeenworkingwithPython,R,andJavaScript.Inmysparetime,Ienjoyworkingonnaturalscienceillustrationsasawayofcombiningmytechnicalandcreativeskills.
BuildingaUnifiedCancerImmunotherapyDataLibrary
9/133:30pm-4:10pm,UnionStationRegencyC
Theintroductionofimmunotherapieshasrevolutionizedthetreatmentofcancerandusheredinacorrespondingexplosionofresearchintocancer,theimmunesystem,andtheirinteraction.InSeptember2018,therewere3,394CancerImmunotherapyagentsbeingtestedinclinicaltrials-a67%increasefromthenumberofagentsbeingtestedoneyearearlier.Withthisexplosioninresearchcomesanexplosionofdataandthechallengeofkeepingtrackofandunifyingpublishedandpublicdatasothatresearchorganizationscanstayontopofthefield.
Toaddressthischallengewe'vecreatedCANDEL,theCancerDataandEvidenceLibrary,adatabasesystemthatleveragesDatomictostoreindividualmolecularmeasurementsandclinicalsamplemetadatainasingleknowledgebase.Datomic'sschemaisexpressive,allowingustomapthevarietyofrepresentationsdifferentorganizationsandresearchersworkwithintoacommonform,aswellastocapturenewconceptsasourknowledgeofmolecularandcancerbiologyevolves.
DatacanbeloadedinthesystemwithoutknowledgeofDatomicinternalsbyusingaconfigurable,data-drivenETLpipelinewedevelopedinClojure.DatascientistscantakeadvantageofthefullpowerofdatalogqueriesdirectlyfromtheirexistingRanalysisenvironment.Furthermore,Datomic'simmutablehistoryensuresanyanalysiswillbereproducible,enablinglargeteamstocollaborateoncomplexanalysisprojectswitheaseandconfidence.
LaceyKitch-gh:ParkerICI-ParkerInstituteforCancerImmunotherapy
LaceyisDirectorofInformaticsatTheParkerInstituteforCancerImmunotherapy(PICI),anonprofitthataimstoacceleratethedevelopmentofImmunotherapiesandgetthemtopatientsfasterthroughenablinggreatresearch.Shehasbeenapplyingdatascienceandmachinelearningtobiologicalandhealth-relateddatasetsformanyyears,includinginherPhDworkatStanford,whereshedevelopednovelalgorithmstoanalyzeopticalbrainimagingdatatolearnaboutmemory.Now,shefocusesonthedevelopmentofdatainfrastructureanditsapplicationintranslationalanalysisatPICI.Asacancersurvivorherself,she'spassionateaboutPICI'scause,andbelievesthatthere'satonthatgoodtechnologycancontributetoacceleratingcuresforcancer.
BenKamphaus@benkamphaus-gh:benkamphaus-Cognitect,Inc.
BenKamphausisasoftwaredeveloperanddatascientistatCognitect.HehasaPhDinGeographyfromSUNYBuffalowherehegothisinitialexposuretointerdisciplinarychallengesindatainfrastructure,includingtheuseofontologies,RDF,andotherapproachestofindingandexpressingcommondatarepresentations.ThesedayshereliesheavilyonClojureandDatomictostaysanewhenmovingdatabetweensystemsandusecontexts.Inhisfreetimehehikes,runs,climbs,writessciencefiction,andproduceselectronicmusicasPatternShift.
ASTRIAGraph:MonitoringGlobalTrafficinSpace!
9/134:30pm-5:20pm,StifelTheatreTheater
ASTRIAGraph:Towardanopen,transparent,andcrowdsourcedspacetrafficandenvironmentmonitoringandawarenesssystem
ÂTheUSdepartmentofdefensetracksapproximately26000residentspaceobjects(RSOs)rangingfromthesizeofasoftballtoaschoolbus.Fromthese,roughly2000areactivelycontrolledandÂallÂelseareeffectivelyspacegarbage.Assuch,theydonottransmittheiridentitiesmakingthemmoredifficulttotrack.Trackingisunderstoodasbothdetectinganobjectandpositivelyidentifyingit.Mostoftheseobjectsaremeasuredaspointdetections(i.e.non-resolved).Astrodynamicsisthesciencethatstudiesmotionofresidentspaceobjects.TherearefourfieldeffectsdrivingRSOmotion:gravitational,radiative,particulates,andgeomagnetic.Ofthese,onlythegravitationalfieldeffectsareindependentoftheRSO'sphysicalcharacteristics.Thequestionthenis,howdowepositivelyidentifyÂallÂoftheRSOsinthepopulationandthusimproveourabilitytopredicttheirbehaviortosatisfyagrowingneedforspacesafety,security,andsustainability?Moreover,howdowedothisinthepresenceofuncertaintydrivenbybothrandomnessandignorance?Dr.ÂJahÂwilldescribecurrentstateofpractice,comparethistostateoftheart,andidentifyscientificandengineeringgapsthatareinneedofbeingsatisfied.HewillalsoprovideasummaryofhisresearchprogramatUTAustinandhowthisfocusareafitsintoalargervisionofrigorousandcomprehensivespacesituationalawarenessandspacetrafficmanagement.
MoribaJah@moribajah-UniversityofTexasatAustin
Dr.ÂMoribaÂJahÂisthedirectorforComputationalAstronauticalSciencesandTechnologies(CAST),agroupwithintheOdenInstituteforComputationalEngineeringandSciences(ICES)atTheUniversityofTexasatAustin.Dr.JahÂcametoUTAustinbywayoftheAirForceResearchLaboratoryandNASA'sJetPropulsionLaboratorypriortothat,wherehewasaSpacecraftNavigatoronseveralMarsmissions.HeisaTEDFellow,aFellowoftheAmericanInstituteofAeronauticsandAstronautics(AIAA),theAmericanAstronauticalSociety(AAS),theRoyalAstronomicalSociety(RAS),theInternationalAssociationfortheAdvancementofSpaceSafety(IAASS),andtheAirForceResearchLaboratory(AFRL).Dr.JahisalsoanelectedmemberoftheInternationalAcademyofAstronautics(IAA)aswellastheInternationalInstituteofSpaceLaw(IISL),hastestifiedtotheUScongressonhiswork,andhasbriefedtheUnitedNationsCommitteeOnpeacefulUsesofOuterSpace(UN-COPUOS)asaninvitedmembertotheUSdelegation.MoredetailsatÂhttps://en.wikipedia.org/wiki/MoribaJah
Unsession:TheDutchCallit"Niksen"
9/137:00pm-8:00pm,UnionStationGrandF
Inatimewheneverythingcanbeoptimizedforproductivity,thereisalwayssomethingyoucouldbedoing.Whatifwemadetimeto"donothing"asanantidotetoallthesesomethings?TheDutchcallit"Niksen."And,asresearcherCarolienHammingisquicktopointout,theydidnotinventordiscoverthis.Ithasbeenathreadthroughoutcultures.Recentresearchindicatesa"purposeless"approachhelpsmanagestress&recoverfromburnout.JennyOdell'srecent"HowtoDoNothing"evenoffersthisasanapproachto"StandApart",wherechoosingto"donothing"becomesquitemeaningful.Asameditatingcoach,Andrewwillprovideanintroductiontopicandofferanoptionalguidedmeditationinatechniqueaptlynamed"DoNothing."(Also,itshouldbeacknowledgedthistopicsitsprecariouslyinatimewhenchoosingto"donothing"comesoutof,orsignals,privilege;though,thisapproachmaybeanactofself-preservation.Thetopicisofferedforconsiderationwithourcontextinmind)
AndrewLenards@alenards
Unsession:GettinginthesaddlewiththePonyprogramminglanguage
9/137:00pm-8:00pm,UnionStationRegencyAB
Ponyisatype-safe,compiled,actor-basedprogramminglanguagethatprovidescompile-timedatasafetyguarantees.Twoofthemostinterestingfeaturesarethereferencecapabilitysystemthatguaranteessafetybycontrollinghowobjectscanbesharedbetweenactors,andtheevent-basedruntime.InthissessionI'llspendsometimeintroducingPonyandthenanswerquestionsandwalkthroughsomecode.BytheendofthesessionparticipantsshouldfeelliketheyknowwhatmakesPonyuniqueandhaveagoodfeelforthekindsofproblemstheymighttacklewithit.
AndrewTurley@casio_juarez
Unsession:Crux-buildyourownbitemporaldatabase
9/137:00pm-8:00pm,UnionStationRegencyC
Crux,asfeaturedearlierthesamedayin"TemporalDatabasesforStreamingArchitectures",isanunbundleddatabasethatprovidesbitemporalversioning,evictionandconsistentad-hocDatalogqueriesontopofKafka.Wewillwalkthroughafewinterplanetarytutorials,liftthehoodonsomeinterestinginternals(Clojure!)andgiveatouroftheuserinterface.Q&Athroughout.
JeremyTaylor
JonPither
LightningTalks
9/138:00pm-9:00pm,UnionStationGrandF
TBD
Unsession:End-to-endTypeSafetywithGraphQL
9/138:00pm-9:00pm,UnionStationRegencyAB
Inmostwebapplications,thefrontendandbackendofaprojectaretiedtogetherbyaRESTAPI,withlittle-to-noguaranteethatchangestothebackendortheAPIwillnotbreakthefrontend.Inthissession,wewillcoverhowGraphQL'sstronglytypedschemacanprotectyoufromeverbreakingyourclient-serverAPIcontractagain,savingyoucountlesshoursonmistakes,typos,andincompatibilities.WewilldemonstratehowyoucanuseGraphQLtoolingtoseamlesslyexploreyourAPIschemawithautocomplete,generatecustomTypeScript/FlowtypesbasedonyourAPIqueries,andintegratethosetypestoachievefulltype-safetyfromthebackendallthewaytothefrontend.Joinusifyou'recuriousaboutGraphQLorifyou'vebeenusingitforyears-nomatterthelevelwe'rehappytochat!
JoãoPeixoto@joaomppeixoto
KristieHoward@kristiehow
Unsession:Healthcaretech,HIPAA,HITRUST,andPHI
9/138:00pm-9:00pm,UnionStationRegencyC
Chataboutthequirksofworkinginhealthcaretech,signingBAAs,audits,querylogging,theincrediblefrustrationofamericanhealthcarepayersystems(insightonothercountriesverywelcome!),flat-files,pharmacyAPIsandpharmacyswitchintegration.IfyouwereatHIMSSwe'dlovetoheararecap/highlights:)
@compiledwrong
Unsession:TheStateOfWirelessRadioTechin2019
9/139:00pm-10:00pm,UnionStationRegencyAB
Let'ssitdownandchataboutvariouskindsofwirelessradiotech(LoRA,LTE(4G/5G/NB),Bluetooth,Sigfox,andstuffIdon'tknowmuchaboutlikeWize),theirpositivesandnegatives,andthecurrentstateofPANs/LPWAN/meshstacksandradiotechforboththetinkererandcommercialuses.ThiswillbelessofadirectedsessionandmoreofabasicoverviewleadingintoaQ&A+roundtablechatforthecuriousand/ortheinterested.
BenLeggett@bleggett
Unsession:Datomic
9/139:00pm-10:00pm,UnionStationRegencyC
ComediscussDatomicwithotherusersand(atleastone)memberoftheDatomicteam.ShortintrotoDatomicandlotsofQ&A
MarshallThompson
BenKamphaus
HowComputersMisunderstandtheWorld
9/149:10am-10:00am,StifelTheatreTheater
AIofferssomanyexcitingpossibilitiesforcampuses,corporations,andcommunities.However,aswelooktothefuture,wemustkeepinmindthattechnologyhasfundamentallimitsandAIisnotamagicbulletthatsolvesallsocialproblems.Inthistalk,authorandprofessorMeredithBroussardlooksattheinnerworkingsandouterlimitsofAIandargueswhyweshouldneverassumethatcomputersgeteverythingright.Makingacaseagainst"techno-chauvisnism"-thebeliefthattechnologyisalwaysthesuperiorsolution-Broussardlooksatwhyself-drivingcarsdon'tactuallyworkandwhyadigital"utopia"willstillharborthesocialproblemsweexperiencetoday.Sheexploreshowunderstandingthelimitsofwhatwecandowithtechnologyallowsustomakebetterchoicesaboutwhatweshoulddowithtechnologytomaketheworldbetterforeveryone.
MeredithBroussard@merbroussard-NewYorkUniversity
MeredithBroussardisanassistantprofessorattheArthurLCarterJournalismInstituteofNewYorkUniversityandtheauthorof"ArtificialUnintelligence:HowComputersMisunderstandtheWorld."Herresearchfocusesonartificialintelligenceininvestigativereporting,withaparticularinterestinusingdataanalysisforsocialgood.SheisanaffiliatefacultymemberattheMooreSloanData
ScienceEnvironmentattheNYUCenterforDataScience,a2019ReynoldsJournalismInstituteFellow,andherworkhasbeensupportedbytheInstituteofMuseumandLibrarySciencesaswellastheTowCenteratColumbiaJournalismSchool.AformerfeatureseditoratthePhiladelphiaInquirer,shehasalsoworkedasasoftwaredeveloperatAT&TBellLabsandtheMITMediaLab.HerfeaturesandessayshaveappearedinTheAtlantic,Harper's,Slate,andotheroutlets.FollowheronTwitterÂ@merbroussardorcontactherviameredithbroussard.com.
ComplexitiesofColorinComputing
9/1410:20am-11:00am,StifelTheatreEncore
Mostofusdon'tquestiontheassumptionswemakeaboutcolorandhowweimplementitinthetechnologywebuild,beyondperhapsaquickcontrastcheck.Thisinattentiontocolorisagravemistake,asthehistoryofcolorincomputingisvibrant,anditsimpactonourlivescanbesignificant!Inthistalk,welearnaboutthedifferencebetweenEuclideandistanceanddelta-ebetweencolors,andhowitcanbreakADAcompliance,crashyourAI,andtrickyourfacialrecognitionsoftwareintothinkingyoudon'texist,allwithoutraisinganeyebrow.
Handwork(knitting,crocheting,cross-stitching,etc.)duringencouraged.
EllenWondra@ln1draw-gh:ln1draw
EllenWondraisa6foot7fire-breathingmountaingoatinthewildsofEasternWashington.I'mjustkidding,butIhadyouworriedforaminutethere,huh?
EllenwasamemberoftheinauguralclassoftheAdaDevelopersAcademyintheFallof2014andithasgivenheranimmenseamountofrespectforthevalueofempathy,community,andhumanityintechnology.Shelikesworkingattheintersectionoftechnologyandart,particularlythefiberarts.Shehascross-stitchedmorethanafewQRcodes,andherfavoritesideprojectisprobablyStitchifier,agemthattakesapictureandreturnsacountedcross-stitchpattern.AskHerAboutHerCats.
CanAndroidsConceiveofElectricSheep
9/1410:20am-11:00am,StifelTheatreOvation
Whatbeganasahackathonprojectturnedintoaminorobsession...
Afewyearsago,Iwroteascriptthatturnsaphotointoananimationofahanddrawingthephoto.AtfirstitwasjustafunhackbutasIironedouttheperfections,Iencounteredseveralinterestingproblemsandfoundsomenovelapplications.HowcanItweaktheanimationtobettersimulatehumanbiomechanics?HowcanIsequencetheanimationtosimulatehumancognition?HowcanIaltertheoriginalimagetosimulatehumancreativity?Howcaninterprettheobjectsandsymbolsintheoriginalimagetosimulatehumaninsight?Canthisdigitalalgorithmicartistinspireusasmuchasit'sanaloghumancounterparts?
InthistalkI'llanswerthosequestionsandmoreasIsharethestorybehindmyattempttocreateanandroidartist.
KwameThomison-gh:Kwamenum86-MagneticInc
KwameisanengineeringleaderwithexperienceatcompanieslikeFacebook,Asana,andMeebo.He'scurrentlyanengineeringleadershipcoachandadvisor.Inhisfreetimehemostenjoysdreamingupideasthatdidn'texistbeforeandmakingthemareality.
Federatedlearning:privatedistributedML
9/1410:20am-11:00am,StifelTheatrePeabodyLounge
Federatedlearningisawaytodomachinelearningwhentrainingdataispartitionedbetweennodesthatareeitherunableorunwillingtoshareit.
Thenodescanbeembeddeddevices,smartphonesorevenlegalentititieslikecompaniesorcountries.TheycanbeunabletosharethedatabecauseofengineeringconstraintslikebandwidthorpowerorlegalbrightlinessuchasHIPAA.Ortheycanbeunwillingtosharethedatabecauseof(verylegitimateandtopical!)concernsaboutthesecurity,commercialexploitationandprivacyofsensitivepersonaldata.
Federatedlearningallowsthenodestocollaboratetotrainamachinelearningmodel,withoutneedingtosharedirectaccesstotheirtrainingdatawitheachotheroracentralauthority.Insteadtheyeachsharepartiallytrainedmodels.
Thistalkwillexplaintheseideasinmoredetail.I'lldescribeaspecificinstanceofafederatedlearningalgorithm(calledfederatedaveraging),andI'llexplainthewaysinwhichtherealworldfullofmaliciousactorsanddistributedsystemscomplicatesthenaivepicture.I'llthentalkabouttheresearchthatisgoingonrightnowtohardensecurity,reducecommunicationcosts,andstrengthenprivacyguarantees.
Thehopeisthat,withfederatedlearning,wenolongerneedtogiveupourprivacyinordertouselife-saving,money-saving,helpfulandfunmachinelearningmodels.
MikeLeeWilliams@mikepqr-gh:williamsmj-Cloudera
MikeLeeWilliamsisanengineeratClouderawhereheworksonmachinelearningandproductsthatmakemachinelearningeasier.WhilegettinghisPhDinastrophysicshespent2%ofhistimeobservingtheheavensinbeautifulfarwestTexas,andtheother98%tryingtofigureouthowtofitstraightlinestodata.HeoncedidapostdocattheMaxPlanckInstituteforExtraterrestrialPhysics,which,amazingly,isarealplace.
BeyondAlt-Text:TrendsinOnlineAccessibility
9/1410:20am-11:00am,StifelTheatrePremier
Ifyou'relikethe2016versionofme,thenyouthinkyouhaveadecenthandleonwebaccessibility.Youputaltattributesonallyourimages(thoughyoudon'tgivemuchthoughttotheactualtext)andyoumakesureyoursitescanbeusedwithakeyboard(exceptforoverlayssometimes).Thenthedaycomeswhenyou'regivena100pageaccessibilityauditfromaclientandadeadlineforallissuestobefixed.Whatishighcontrastmode,youaskyourself,andwhydoesitmatteriftheselinksareimplementedinalist?
Thefactisthatwetakeforgrantedthatpeopleareabletousethesiteswecreate.Asgovernmentspasslegislationenforcingaccessiblestandards,thewebisplayingcatchuptothephysicalspacesweuseeveryday.Userinterfacesarebecomingmoresophisticated,andit'seasyfordevelopers&designerswithoutdisabilitiestooverlookthesimplethingsthatmanyrelyontomakesenseofyoursite.
InthispresentationI'llsharesomeofthehardlessonsI'velearnedoverthepastfewyears,frombothadevelopmentandaprojectmanagementperspective.Topicsdiscussedwillincludeanoverviewofcommonaccessibilityproblems,toolsIusetovalidateaccessibilityissues,andbestpracticesfortrainingyourteam.
IanForrest@ianforr-BiblioCommons
IanisanEngineeringManagerbasedoutofToronto.CurrentlyheworksatBiblioCommons,whereheleadsthedevelopmentofamulti-siteWordPressapplicationusedbypubliclibrariesinovertwentycitiesincludingSt.Louis,Chicago,andBoston.Hehasalsohadsomesuccessfulsideprojectsincludinggradient-animator.com,aCSSgeneratorforanimatedgradients.Inhisfreetimeheenjoyswatchingoldmoviesandhopelesslycheeringforlocalsportsteams.
HowWeTellStoriesInCode
9/1410:20am-11:00am,StifelTheatreTheater
Ifyouyouasktheexpertshowtowrite'good'software,theywilllikelyemphasizeprecisionandpredictability.Theymayrecommendgreattoolslikefunctionalprogramming,immutability,testing,andtypesystems-allofwhichaimtomakesoftwaremorereliable.Butsoftwaremustinteractwiththerealworldtobeuseful,andtherealworldalwayspresentsscenariossoftwaredoesnotaccountfor.Youcan'twritegoodsoftwarebytryingtoarchitectunpredictabilityoutofyourprogram.Towritegoodworkingsoftware,youmustbuildsystemsthatworkwelleventhoughtheyareunpredictableandincomplete.Yourcodehastodescribeandsolvearealworldproblemwithoutaccountingforalldetailsoftheproblem.Goodsoftwaretellsanarrative,butlikeallnarrativesleavesthingsout.Maybegoodprogrammersarereallyjustgoodstorytellers.
Storytellingisaprocessthatwedoallthetime,includingwithourpersonalhistories.Storytellingbothapproximatesthetruthandinfluencesourreality-thestorieswetellaffecthowwerememberthepastandbehaveinthefuture.Thishasinterestingimplicationsforwritingsoftware-thesystemswedesignaffectandchangethenatureoftherealproblemsthey'redesignedtoaddress.Andthestorywetellincodeisdirectedasmuchtootherprogrammersasitistothemachinethecoderunson.Recognizingthemessy,veryhumannatureofwritingsoftwareteachesustotrustourintuition,andmakesusbetterprogrammers.
HannahHoward@techgirlwonder-gh:hannahhoward
HannahHowardisaseniordeveloperandtechgeneralistwithover15yearsexperienceinprogrammingandothertechnicalfields.Priortoprogramming,Hannahworkedfor10yearsinthenon-profitsectorinLosAngeles,specializinginLGBTadvocacyandcommunityorganizing.Hannahreturnedtocodingin2012,andbringsherpassionandexperiencefromcommunityorganizingtohelpingnewprogrammersgetuptospeedontechnicaltopics.
Securingcloudapplicationsandtheircredentials
9/1410:20am-11:00am,UnionStationGrandF
Withtheriseofcontainerizedcloud-basedinfrastructure,securityinthesearchitectureshasbeenanareathathasbeenoftendeferredtodevelopersand"shadowIT"becausethetoolsandtechnologiesusedpreviouslyhavehavenotbeenasquicktocatchup.Whiledevelopersgenerallymakereasonabledecisionsaboutsecuringworkloads,pushesforhighervelocityendupcausingsimple
implementationmisconfigurationsthatexponentiallyincreasetheattacksurface.Attendeeswilllearncommoncontainersecurityproblemsaswellashowtoincreasesecurityofyourappsbothonthecontainerandschedulerlevel.Wewillalsocoversomeofthenext-gentoolingandseewherelong-termsecurityinitiativesareheadinginthisspace.
SrdjanGrubor@sgnn7-gh:sgnn7-CyberArk
SrdjanGruborisanR&DsoftwareengineerwhoworksatCyberArkonnext-gensystemstosecurecloud-basedsystems.Srdjanistheauthorof"DeploymentwithDocker"book,wasoneofthefirstpeopletoreceiveaDockerCertifiedAssociatecertification,andhasworkedonLinuxsystemsat-scaleforoveradecade.Heenjoysbreakingthingsjusttoseehowtheywork,tinkering,andsolvingchallengingproblems.
MinimalistPianoForever
9/1410:20am-11:00am,UnionStationRegencyAB
FrenchcomposerErikSatiewastheforefatherofavant-gardemusic,thefounderofaone-manoccultsect,andtheinventorofentirelynewgenresforhishauntinglyrepetitive,minimalistpianoworks.Personally,Icouldlistentohis"Gnossiennes"forever,soIcreatedawaytodoso,usingdigitalsheetmusic,markovchains,andbrowser-basedmiditocreateagenerative,unendingGnossienne.Thistalkwillexplorethestrategies,algorithms,andlibrariesIusedtogofrom19thand20thcenturypianomusictogenerativeaudioandon-the-flysheetmusicinthebrowser.Itfeaturesforaysintotopicsrangingfromhowtomakecomputergeneratedmusicsoundmorenatural,toErikSatie'sverypeculiarlife.
MouseReeve@tripofmice-gh:mouse-reeve-InternetArchive
MouseisasoftwareengineerattheInternetArchive,workingonpreservingthewebandmakingitfreelyavailabletoanyoneontheinternet.Theyhaveanacademicbackgroundinanthropologyandmathematics,withstronginterestsinlinguistics,generativeart,andhistoricaloccultism.Mousecreatedtheinternet'sbestsocialnetworkwherenooneisallowedtousethelettere,anddabblesinveganpatisserie,socialjustice,andlookingatflowers.Theycareverydeeplyaboutnonsense.
Whydotreeensembleswork?
9/1410:20am-11:00am,UnionStationRegencyC
Ensemblesofdecisiontrees(e.g.,therandomforestandAdaBoostalgorithms)arepowerfulandwell-knownmethodsofclassificationandregression.Thistalkwillsurveyworkaimedatunderstandingthestatisticalpropertiesofdecisiontreeensembles,withthegoalofexplainingwhytheywork.Aftersketchingthealgorithms,wewillgiveaninitialexplanationfortheireffectivenessviagenericarguments(bias-variancedecomposition,Hoeffding'sinequality),thenproceedtomoredetailedtopics(theinterpretationofrandomforestsaskernelmachines,theroleofthemargin,interpolation).Theaudienceisexpectedtohavesomeexperiencewithsupervisedlearningandstatisticalarguments.
JoeRoss@robusteza-SignalFx
JoeRossholdsaPhDinmathematicsfromColumbiaUniversityandwasaresearcherandinstructorinpuremathematics,mostrecentlyattheUniversityofSouthernCalifornia.Hehasgivenmorethan20talksabouthisresearchatconferencesanduniversitiesthroughouttheworld(Germany,Japan,Turkey,USA).Hehasalsobeentheprimarylecturerformanyundergraduateandgraduatemathcourses,andhasgivencountlessinformalseminars.Hehas9publicationsinpeer-reviewedmathematicsjournals.Joehasworkedasadatascientistatmachinelearning/analyticsstartupsforfiveyears;inhiscurrentrole,hefocusesonavarietyoftimeseries(anomalydetection,forecasting,correlation)andsamplingproblemsthatariseinmonitoring.
CompactingtheUncompactable
9/1411:20am-12:00pm,StifelTheatreEncore
ProgramswritteninC/C++,cansufferfromseriousmemoryfragmentation,leadingtolowutilizationofmemory,degradedperformance,andapplicationfailureduetomemoryexhaustion.ThisproblemextendstolanguageslikeRubyandPython,wherethestandardinterpretersfortheselanguagesarethemselvesCprograms.ThistalkintroducesMesh,aplug-inreplacementformallocthat,forthefirsttime,eliminatesfragmentationinunmodifiedC/C++applicationsthroughcompaction.Akeychallengeisthat,unlikeingarbage-collectedenvironments,theaddressesofallocatedobjectsinCandC++aredirectlyexposedtoprogrammers,andapplicationsmaydothingslikestashaddressesinintegersorstoreflagsinthelowbitsofalignedaddresses.Thishostileenvironmentmakesitimpossibletosafelyrelocateobjects,astheruntimecannotpreciselylocateandupdatepointers.Meshcombinesnovelrandomizedalgorithmswithwidely-supportedvirtualmemoryoperationstoprovablyreducefragmentation,breakinglong-establishedworst-caseboundsonmemoryefficiencywithhighprobability.Meshgenerallymatchestheruntimeperformanceofstate-of-theartmemoryallocatorswhilereducingmemoryconsumptionandeliminatingpathologicalcases;inparticular,MeshreducesthememoryofconsumptionofFirefoxby16%andRedisby39%.ThereareeffortsunderwaytoincorporateMesh'sapproachtoeliminatefragmentationintoexistingallocators,likejemalloc.
BobbyPowers@lilbobbypowers-gh:bpowers-UniversityofMassachusettsAmherst
BobbyisaPhDcandidateinthePLASMALabatUMassAmherst,andasoftwareengineeratStripe.Hisinterestsspansystemsandprogramminglanguages,withafocusonmakingexistingsoftwaremoreefficient,moresecure,andusableinnewcontexts.
NetworkingandGo:AnEngineer'sJourney
9/1411:20am-12:00pm,StifelTheatreOvation
MyforayintoGobeganafewyearsagowhenIstartedworkingatDigitalOcean.WhilebuildinganabstractionlayerontopofKubernetesandfamiliarizingmyselfwiththelanguage,Ibegantoloveit.Syntacticallysimple,withamazingconcurrencyprimitivesandawonderfulcommunity,Gowasanexcellentchoiceforacloud-hostingcompanywithavarietyoflow-level,server-sidemicroservices.
Inthelastyear,however,I'vejoinedthesoftware-definednetworkingteamandlearnedofanotherapplicationofGo;networkingservices.ThenetworkingteamatDigitalOceanusesGoforavarietyofpurposes-fromDHCPserverstoIPaddressmanagementservices..toevenwrappersaroundvirtualswitchtooling.Intrigued,IdecidedtoalsoinvestigatehowGocouldbeusedtobuildotherservicessuchasportscannersandload-balancers.
ThissessionwillhighlightmynetworkingjourneyviaGo.Iwilldiscussusefulpackages,keylearnings,andevenstrugglesfacedwhilebuildingavarietynetworkingserviceswithinandoutsideofDigitalOcean.Iwilldiscussbothrelevantpackageswithinthestandardlibraryandopensourcepackagesusedtoimplementkeynetworkprotocols.Asaresult,listenerswillgainanunderstandingofhowtospecificallyleverageGofortheirownnetworkingneeds.
SnehaInguva@snehainguva-gh:si74-DigitalOcean
SnehaInguvaisasoftwareengineerinthenetworkingteamatDigitalOcean.Sheenjoysbuildingcloudproductsbydayanddebuggingominouscontext-cancellederrorsbynight.Interestingly,herjourneythroughthisfantasticalfieldhastakenherfromcasinogamingto3dprintingstartups.Inhersparetime,shetakeshighqualityfashionphotographsofcatsinnature.
BuildingHaskellProgramswithFusedEffects
9/1411:20am-12:00pm,StifelTheatrePeabodyLounge
Haskellisapurelyfunctionalprogramminglanguage:bydefault,HaskellfunctionsdonotcausesideeffectssuchassystemI/O,nondeterminism,orexceptionhandling.Assuch,Haskellprogramsaregenerallyexpressedintermsofmonadtransformers,whichprovidethefacilitytocomposedifferentsideeffectsintoasingleinterfacepowerfulenoughtoexpresstheprogrammer'sneeds.Themonadtransformerlibrary,mtl,ismatureandpowerful,butcomplicatesandinsomecasesconstrainstheconstructionandgeneralizationofuser-specifiedmonads.
Aclassoflibrariesknownas'effectsystems'haveemergedinanattempttosolvetheproblemsassociatedwithmonadtransformers.Effectsystemsprovideacompositionalapproachtoprogramconstruction,generallyusingasinglemonadictypespecialzedwithaprogrammer-providedlistofcapabilities.Effectsystemsaremorepowerfulandflexiblethanmonadtransformers,buttheiradoptioninindustryhasbeenminimal,duebothtotheirincompatiblitywithmtlandtheirhistoricallypoorperformance.
Thenewfused-effectslibrarychangesthestatusquo:itprovidesanextensibleandflexiblevocabularyforprogramconstruction,yieldingamoreexpressiveinterfacethanmtlwithoutsacrificinganyperformancecharacteristics.I'lldescribethehistoryofeffectsystems,outlinethetradeoffsassociatedwithprogrammingwitheffects,anddemonstratetheuseoffused-effectsinpractice.
PatrickThomson@importantshock-gh:patrickt-GitHub,Inc.
PatrickisaseniorengineeronGitHub'sSemanticCodeteam,buildingsystemstounderstandandanalyzethecorpusofcodeonGitHub.Heenjoyspeacefulcountrysidewalksandloudrapshows.
FunctionalDevOpsinaDysfunctionalWorld
9/1411:20am-12:00pm,StifelTheatrePremier
Nixisapackagemanager,andaprogramminglanguageforwritingpackagesinthatpackagemanager,thatenablesreliableandreproduciblesoftwaredevelopmentworkflows,e.g.areproducibledevenvironment,apackagethatinstallscorrectlyeverytime,andevenpain-freeservicedeployments.InthissessionIplantodemonstratepackaginganddeployingsoftwarewithNix,afterwhichIwilltakequestionsand/orstartadiscussion.BytheendoftheunsessionparticipantsshouldhaveaworkingunderstandingofwhatproblemsNixcansolveandhowtointegrateitintotheirsoftwaredevelopmentworkflows.
VaibhavSagar@vbhvsgr-ObsidianSystems
Vaibhav(@vbhvsgr)usedtowritewebapplicationsforaliving.Hestilldoes,butheusedto,too.Whenhe'snotdoingthatheyellsabout
functionalprogrammingandpackagemanagersonTwitter,re-racksbumperplatesafterusingthem,andtrawlstheinternetforthefinestspacedisco.
CorrectnessproofsofdistributedsystemswithIsabelle
9/1411:20am-12:00pm,StifelTheatreTheater
Testingsystemsisgreat,buttestscanonlyexploreafinitesetofinputsandbehaviors.Manyrealsystems,especiallydistributedsystems,haveapotentiallyinfinitestatespace.Ifyouwanttobesurethataprogramdoestherightthinginallpossiblesituations,testingisnotsufficient:youneedproof.Onlymathematicalproof,e.g.byinduction,cancoveraninfinitestatespace.
Pen-and-paperproofsarewellestablishedinmathematics,buttheyneedtobelaboriouslycheckedbyhand,andhumanssometimesmakemistakes.Automatedtheoremproversandcomputerizedproofassistantscanhelphere.ThistalkintroducesIsabelle/HOL,aninteractiveproofassistantthatcanbeusedtoformallyprovethecorrectnessofalgorithms.ItissomewhatlikeaprogramminglanguageandREPLforproofs.
InthistalkwewillexplorehowIsabellecanbeusedtoanalyzealgorithmsfordistributedsystems,andprovethemcorrect.Wewillworkthroughsomeexampleproblemsinlivedemos,andproverealtheoremsaboutsomesimplealgorithms.Proofassistantsstillhaveaprettysteeplearningcurve,andthistalkwon'tbeabletoteachyoueverything,butyouwillgetasenseofthestyleofreasoning,andmaybeyouwillbetemptedtotryitforyourself.
MartinKleppmann@martinkl-gh:ept-UniversityofCambridge
DrMartinKleppmannisaresearcherindistributedsystemsattheUniversityofCambridge,andauthoroftheacclaimed"DesigningData-IntensiveApplications"(O'ReillyMedia,2017).Hemainlyworksoncollaborationsoftware,CRDTs,andformalverificationofdistributedalgorithms.PreviouslyhewasasoftwareengineerandentrepreneuratInternetcompaniesincludingLinkedInandRapportive,whereheworkedonlarge-scaledatainfrastructure.
Probabilisticscriptsforautomatingcommon-sensetasks
9/1411:20am-12:00pm,UnionStationGrandF
Asengineers,weloveautomatingtedioustasks.Butwhenthosetasksrequirecommon-sensereasoning,automationcanbedifficult.Consider,forexample,cleaningamessydataset-fulloftypos,NULLvalues,numbersinthewrongunits,andotherproblems.Peoplehavelittletroublefixingtheseerrorsbyhand,butitcanbedifficulttoexpresstherulesfordoingsoprogrammatically.
Inthistalk,I'llintroduceanewdeclarative-programmingapproachforautomatingcommon-sensereasoningtasks:probabilisticscripting.Probabilisticscriptsencode(possiblyuncertain)domainknowledgedeclaratively,andleavethecompilertosynthesizeanefficientinferencealgorithmthatwillsolvethetaskathand.Thisisallmadepossiblebyrecentadvancesinthefieldofprobabilisticprogramming-inparticular,programmableinferenceengines.
Iwilldemonstratehowthistechniquecanbeusedtodesignandimplementascriptinglanguageforautomatingreal-worlddata-cleaningtasks,whichachievesstate-of-the-artaccuracyondata-cleaningbenchmarks.Morebroadly,attendeeswillcomeawaywithasenseofhowprobabilisticprogrammingcanbeusedtobringcommon-sensereasoningtotheautomationofallsortsoftasks.
AlexanderLew-gh:alex-lew-MITProbabilisticComputingProject
AlexLewisaPh.D.studentatMIT'sProbabilisticComputingProject,andaleadresearcherforMetaprob,anopen-sourceprobabilisticprogramminglanguageembeddedinClojure(Script).Heaimstobuildtoolsthatempowereveryonetouseprobabilisticmodelingandinferencetosolveproblemscreatively.BeforecomingtoMIT,Alexdesignedandtaughtafour-yearhigh-schoolcomputersciencecurriculumattheCommonwealthSchoolinBoston.AnativeofDurham,NC,healsoreturnshomeeachsummertoteachattheDukeMachineLearningSummerSchool(andspendtimewithhisfamilyandtheirdogs!).
FromVideoGamestoFashion:aMachineLearningJourney
9/1411:20am-12:00pm,UnionStationRegencyAB
Gamingisarapidlygrowingindustry.Playersgenerateextremelyrichdatasetsthatrecordeachandeveryin-gameaction,socialinteraction,advieworpurchase.Becauseoftheconstantandcontinuedrelationshipbetweenplayerandgame,playerbehavioraldataconstitutesauniquesourceofinformationonhumanbehavior:asimplifieddynamicsystemthatreplicatesreallife.Â
Fashiongoesbeyondabasicneed:itisawaytoexpressourownindividuality.Thefashionindustryisoneofthelargestintheworld,representing2percentofglobalGDP,andhasstrongsocialandeconomicimplicationsworldwide.Vastamountsofdataareconstantlygenerated,rangingfromcustomerbehaviortocomplexsupply-chainsystems.Successinthefashionindustryreliesoncarefullybalancingartandscience.
InthistalkIwilldiscussthecommonchallengestobothindustriesandtheemergencyofmodelsthatscaletoreallybigdatasetsandmanagetocapturevariouskindsofconsumertastesandbehaviors.Becauseoftheimportanceofthetemporaldimension,modelsbasedonCNN,LSTMorDeepARplayacrucialroletopushbackthisnewtechnologicalfrontier.Iwillreviewhowthesetechniquescanbeusedtocreatecustomizedeventsandprovidepersonalizedpredictions.
AfricaPerianez@aperianez-Inditex
ÃfricaPeriáñezistheChiefAnalyticsOfficeratInditex.Ãfricaleadstheresearch,applicationandstrategicdevelopmentofdatascienceacrosstheorganization.Hergoalistobuildrobustmachinelearningproductstochampionfashionretailingexperience.
BeforelandinginfashionindustryÃfricafoundedYokozunaData,anAIcompanyheadquarteredinTokyothatbuiltapioneermachinelearningplatformtopredictvideogameplayerbehavior.ÃfricaholdsaPhDinMathematicsfromtheUniversityofReading,andMSc'sinStringTheoryandTheoreticalPhysicsfromCERNandtheAutonomousUniversityofMadrid.
ÃfricahasbeenMarieCurieEUresearchfellowatCERNandscientistatRIKEN,Japan(usingKcomputer)andtheGermanWeatherService(workinginsatellitedataassimilation).Sheisco-authorofmultiplepeer-reviewedarticlesandregularinvitedspeakerininternationalconferences.ÃfricahassolidindustryexperienceatSPSSInc.,IBMandSiliconStudio.HerworkhasbeenfeaturedinmediasuchasBloomberg,Quartz,Nikkei,ElPaÃs,SINC,TheEconomicTimesandTheJapanTimes.Â
AbetterstoryforKubernetessecrets
9/1411:20am-12:00pm,UnionStationRegencyC
SecretsareakeypillarofKubernetes,butanyonewithaccesstoetcdcanalltheplaintextvalues!
AttendeeswilllearntechniquesforsecuringKubernetessecretsincludingencryption,KMSplugins,andtoolslikeHashiCorpVault,andthetradeoffsofeachapproachtobettersecuretheirclusters.
SethVargo@sethvargo-gh:sethvargo-Google
SethVargoisanengineeratGoogleCloud.PreviouslyheworkedatHashiCorp,ChefSoftware,CustomInk,andsomePittsburgh-basedstartups.HeistheauthorofLearningChefandispassionateaboutreducinginequalityintechnology.Whenheisnotwriting,workingonopensource,teaching,orspeakingatconferences,Sethadvisesnon-profits.
SaturdayLunch
9/1412:00pm-1:30pm,StifelTheatreTheater
LunchisprovidedatboththeStifelTheatreANDatUnionStation.Itisdifficulttosizethemealsappropriatelyaswearesplitbetweenthetwovenues,sowe'dpreferthatyoustayatthesamevenueasthesessionyouattendedjustpriortolunch.
StifelTheatrelunch:
Option1:GrilledChickenCaesarSalad-crispromaine,parmesanandgarliccroutons,classicCaesardressingOption2:RoastLoinofPorkSandwich-Bavarianpretzelbread,tarragonmayo,sweetpeppers,arugulaVegan(servedinKielClub):Chik'nSalad-organicgrainbread,mashedchickpeas,dicedonion,carrots,relish,veganmayo,salt,pepper,paprikaGlutenfree(servedinKielClub):ChickenCaesar,nocroutons,orPorkSandwichSides-orangeorapple,cookie
UnionStationlunch(veganoptionsonbuffet,willbelabeled):
CaesarSalad,ShavedParmesan,Crouton,TraditionalDressing(deconstructed,balsamicoption)IndividualTomato&MozzarellaSaladCrackedPepper,ExtraVirginOliveOilOlive,Artichoke&SalamiSaladAssortedGourmetPizzas(GFandVegetarianOptions)ToastedRavioliwithMarinaraVegetableRavioli,LightTomatoBroth,Basil,RoastedGarlicOilPennewithRockShrimp,Peas,Pesto(GFPasta)ShavedParmesan&Tomato,GarlicBreadSticksAlmondBiscotti,Tiramisu
ExplainableAI:theapexofhumanandmachinelearning
9/141:30pm-2:10pm,StifelTheatreEncore
BlackBoxAItechnologieslikeDeepLearninghaveseengreatsuccessindomainslikeaddelivery,speechrecognition,andimageclassification;andhaveevendefeatedtheworld'sbesthumanplayersinGo,Starcraft,andDOTA.Asaresult,adoptionofthesetechnologieshasskyrocketed.ButasemploymentofBlackBoxAIincreasesinsafety-intensiveandscientificdomains,wearelearninghardlessonsabouttheirlimitations:theygowrongunexpectedlyandaredifficulttodiagnose.
Fromthesefailures,anewtrendof"ExplainableAI"hasemerged.TheseareAItechnologiesdesignedtobeintuitiveandunderstandabletotheirhumanuserswhilemaintainingthepowerandexpressivenessofBlackBoxAI.
InthistalkwewilldiscussexplainableAI:whatitis,whenandwhyit'sneeded,andhowtobuildit.Wewillexplorethefundamentaldifferencesbetweenhumanandmachinelearning,anddiscussresearchattheapexofcomputationandcognitionthathasleadtomachinesthatarenotonlyintuitiveandunderstandabletodatascientists,butcanefficientlycommunicatetheirknowledgetoanyonebyexploitinghumans'innatesociallearningcapabilities.
BaxterEaves-gh:[email protected]
BaxterEaves,PhDiscurrentlyco-founderofRedpoll,acompanybuildinghumanisticAItohelpdrivescience.HereceivedhisPhDinexperimentalpsychologyfromtheUniversityofLouisvillewherehebuiltmachinesthatlearntotrustanddistrust.SincethenhehasworkedonprobabilisticprogramminglanguagesatMIT,MachineTeachingatRutgers,GenomicSelectionandMonsanto,andhasbeeninvolvedwiththeDARPAXDATAandPPAMLprojects.Ithasbeenhislifelonggoaltobuild,andbedestroyedby,thefirstsentientmachine.
ParserParserCombinatorsforProgramTransformation
9/141:30pm-2:10pm,StifelTheatreOvation
Multi-languagetoolsupportforsyntaxtransformationishardduetoheterogeneousrepresentationsinsyntaxandabstractsyntaxtrees(ASTs).Regex-basedsearch-and-replacefallsshortofrecognizingsyntaxthatfundamentallydelineatestreedatastructures.Recentapproachesdevelopnewstrategiesthatovercomethelimitationsofregexmatchingbutremainunderdevelopedforeasilychangingcode.Ourworkgoesonestepfurther,focusingontheproblemofenablinglightweightprogramtransformationineverylanguageforeveryprogrammer.Weshowthattheproblemcanbedecomposedwhere(1)acommongrammarexpressesthecentralcontext-freelanguagepropertiessharedbymanycontemporarylanguages(e.g.,balancedparentheses)and(2)openextensionpointsinthegrammarcustomizessyntaxhandling(e.g.,forlanguage-specificcomments)withsmallerparsers.WeintroduceParserParserCombinators(PPCs),ourkeymechanismimplementingtheseideas.PPCsareparsercombinatorsthatproduceparsersfromuser-suppliedpatterns.Generatedparsersrundirectlyonprogramsourcetomatchsyntaxofinterest(wedon'tdefineoruseanyAST),therebyliftingsyntaxrewritingtoamodularly-definedparsingproblem.Wesharelarge-scaleresultsfromrewritingcodeacross12languages(Go,Rust,Scala,andElmtonamebutafew)fortop-100mostpopularGitHubrepositories(perlanguage).Weshowover50syntacticchangesmergedinto40+oftheseprojectsusingourtool,andgiveademo.
RijnardvanTonder@rvtond-gh:rvantonder-CarnegieMellonUniversityandSourcegraph
RijnardisaPhDcandidateatCarnegieMellonUniversityandapart-timesoftwareengineeratSourcegraph.HisresearchinterestisintheoverlapofAutomatedProgramRepair,ProgramTransformation,andProgramAnalysis,withanemphasisonbringingnewadvancesinthisareatopractice.RijnardisbornSouthAfricanandholdsaMaster'sandBachelor'sfromStellenboschUniversity.
ComputerVisionandNLPforMulti-TaskFashionModeling
9/141:30pm-2:10pm,StifelTheatrePeabodyLounge
Shoprunneraggregatesmillionsofproductsfrom140retailerswhichrepresentthousandsofbrands.InordertomaketheseproductsfindableandsearchablebyusersitisimportantforShoprunnertobeabletostandardizetheattributes(style,color,patternetc)ofthesemillionsofproducts.Evenafterdefiningwhatattributestomodel,choosingthebestwaytopredictattributesisdifficultbecauseeveryproductcanberepresentedinavarietyofformssuchasimages,productdescription,title,andbrandname.Thesedifferentdatarepresentationseachhavetheirstrengthsandweaknesses.Imagesencodeinformationsuchascolorandpatternwellwhileotherattributesrelatedtolengthandcutmaybewellcapturedintextdescriptions.Thissessionwillgothroughthemulti-tasklearningensemblethattheDataScienceteamatShoprunnerhasbuiltusingbothcustommulti-taskCNNsforimagesandfine-tunedBertmodelfortextclassificationinPytorchforattributemodeling.
MichaelSugimura@sugichan014-gh:sugi-chan-Shoprunner
IamaSeniorDataScientistwhofocusesoncomputervisionapplicationsatShoprunner.InparticularIworkonbuildingattributemodelsusingmulti-taskensemblesandbuildingdetectionandsegmentationmodelsappliedtothee-commercespace.Whennotworkingondatascienceatwork,Michaelisusuallyworkingonsidedatascienceprojectsandbloggingaboutthem,doingmartialarts(WingChunKungFu,MusoJikidenEishinRyuIaijutsu,Kendo),orcooking.
EverythingYouWantedtoKnowAboutDistributedTracing
9/141:30pm-2:10pm,StifelTheatrePremier
Intheageofmicroservices,understandinghowapplicationsareexecutinginahighlydistributedenvironmentcanbecomplicated.Lookingatlogfilesonlygivesasnapshotofthewholestoryandlookingatasingleserviceinisolationsimplydoesnotgiveenoughinformation.Eachserviceisjustonesideofabiggerstory.Distributedtracinghasemergedasaninvaluabletechniquethatsucceedsinsummarizingallsidesofthestoryintoasharedtimeline.Yetdeployingitcanbequitechallenging,especiallyinthelargescale,polyglotenvironmentsofmoderncompaniesthatmixtogethermanydifferenttechnologies.Duringthissession,wewilltakealookatpatternsandmeanstoimplementTracingforservices.Afterintroducingthebasicconceptswewillcoverhowthetracingmodelworks,andhowtosafelyuseitinproductiontotroubleshootanddiagnoseissues.
HungaiKevinAmuhinda@Hungai-gh:hungaikev
HungaiisaDeveloperSupportEngineeratLightbendandco-founderofasoftwareautomationcompany(Djuaji).Hehasmorethan6yearsofexperienceinsoftwareengineering,consultingandtrainingandisanactiveopensourcecontributorandorganizerofameetupinNairobi.HeisadatanerdanddigitalentrepreneurbasedinNairobi,Kenya.HungaienjoysworkingwithGo,Scala,Rust,Kafka,Kubernetes,andSpark.
BetterIntegrationTestsforPerformanceMonitoring
9/141:30pm-2:10pm,StifelTheatreTheater
Inlate2017,Slack'slargestcustomerswereplaguedwithrelentlessperformance-relatedoutages.Ourmonolithic,spaghetticodebasewasincreasinglydifficulttoreasonabout;small,innocuouschangesmightaccidentallycauseacascadeofregressions.Afewconcernedengineersteameduptobuildatooldetectingandpreventingthesechangesfromreachingproduction.Knownas"Slerf"(for"Sl"ackP"erf"ormance),thesimplesystemrunsatopSlack'sexistingunittestingframeworkandalertsbackendengineersearlytoanypotentialregressionstheircodemightintroduce.Learnhowyoucanextendyourowntestingframeworkandimproveboththeconfidenceofyourengineersandstabilityofyourproductwhilemaintainingahighdevelopmentvelocity!
MaudeLemaire@qcmaude-Slack
MaudeisastaffengineeratSlackTechnologies,workingtoscaleboththeproductandthebackendengineeringteamtosupportsomeoftheworld'slargestcompanies.Shespendsmostofhertimechasingdownpeoplemakingnetworkcallsinaloop,refactoringunwieldychunksofcode,andtweakingdevelopertools.
PriortomovingtotheBayArea,Maudewasafront-endengineeratRenttheRunwayinNewYorkCity.SheearnedaBSc.inSoftwareEngineeringfromMcGillUniversityinMontréal,QC.Inhersparetime,she'sanavidbakerwhoenjoystakesillypicturesofherhusband.
BuildinganOpenSourceArtificialPancreas
9/141:30pm-2:10pm,UnionStationGrandF
Haveyoueverthoughtaboutwhatopensourcesoftwareorhardwarecouldachieve?Whatifitcouldhelpimprovepeople'slivesbysolvingsomeoftheirhealthproblems?
Afterthemedicaltechindustrykeptpromisingasystemtohelpautomaticallymanageinsulinfortype1diabeticpeopleandneverdelivering,somepeoplegottogethertofindwaystodoitwiththetechtheyalreadyhad.Overthepastfewyears,a"closed-loop"systemhasbeendevelopedtoalgorithmicallyregulatepeople'sbloodsugars.Afterreverseengineeringbluetoothsensorsand915MHzinsulinpumps,thesystembecamepossible.Asadiabetic,IalsobuiltthissystemandsawmysugarvaluesstabilizemuchmorethanIcouldeverachievedoingitmanuallymyself.NowI'mworkingoncontributingbacktotheprojectsaswell.
Iwanttotalkaboutthissystem,fromatechnicalsideaswellasapersonalside.I'lltalkaboutOpenAPS(theopenartificialpancreassystem)andhowitworks,whatproblemsitsolves,anditssafetyandsecurityconcerns.Ialsowanttoshowhowit'shelpedme,andwhatthismeansformyhealthnowandinthefuture.Iultimatelywanttoshowhowwe,assoftwaredevelopers,canchangepeople'slivesthroughthecodewewrite.
SarahWithee@geekygirlsarah-gh:geekygirlsarah
SarahWitheeisapolyglotsoftwareengineer,internationalpublicspeaker,andhardware/robottinkererinPittsburgh,PA.ShealsoiscurrentlytheDirectorofProgrammingforAbstractionsconference.She'shadapassionfortechsincewritingherfirstprograminfirstgrade.Shecaptivatesaudienceswithbothpopularandpowerfultechnicalandanecdotaltalks.Shegivesworkshopstoteachprogrammingandhardwarebuildingtowomenintech,aswellastostudentsofallages.She'salsomentoredyouthroboticsteamstochampionships,organized6conferences,beenonaGoogleYearinSearchvideo,andstartedtheviralhashtag#SpeakerConfessions.
Unison:anewdistributedprogramminglanguage
9/141:30pm-2:10pm,UnionStationRegencyAB
Unisonisanopensourcefunctionalprogramminglanguagewithspecialsupportforbuildingdistributed,elasticsystems.Itbeganasanexperiment:rethinkallaspectsoftheprogrammingexperience,includingthecorelanguage,runtime,tooling,aswellascodeversioningandpublishing,andthendowhateverisnecessarytoeliminateneedlesscomplexityandmakebuildingsoftwareonceagaindelightful,orattheveryleast,reasonable.
We'reusedtothinkingofaprogramasathingthatdescribeswhatasingleOSprocesswilldo,andthenusingaseparatelayeroftechnologiesoutsideofourprogramminglanguagesto"configure"manyseparateprogramsintoasingledistributed,elastic"system".Thisgetscomplicated.ThecorelanguageofUnisonstartswiththepremisethatnomatterhowmanynodesacomputationoccupies,itshouldbeexpressibleviaasingleprogram,notmanyseparateprograms.Unisonprogramscandescribetheirowndeployment,elasticallyscaleandorchestratethemselves,anddeploythemselvesinparallelontoanynumberofnodesforexecution.
ThistalkintroducestheUnisonlanguageanditstoolingandshowswhatitcanbeliketoprogramsystemsofanysizewiththismodelofcomputing.
PaulChiusano@pchiusano-gh:pchiusano-UnisonComputing
PaulChiusanostartedtheresearchthatledtotheUnisonlanguageandisacofounderofUnisonComputing,apublicbenefitcorp.HehasoveradecadeofexperiencewithpurelyfunctionalprogramminginHaskellandScalaandcoauthoredthebookFunctionalProgramminginScala.HelivesandworksinSomerville,MA.
CSSAlgorithms
9/141:30pm-2:10pm,UnionStationRegencyC
Wait...CSSalgorithms?CanonereallywritealgorithmsinCSS?Absolutely.
ThistalkisallaboutCSSalgorithms:whattheyare,howtowritethem,andthepotentialimpactsofawordlike"algorithm"whenitcomestoboththeCSSinyourcode-basesandCSSinthelargercontextofwebdevelopment.
LaraSchenck@laras126-gh:laras126-PenskeMediaCorporation
LaraSchenckisfascinatedbytheprofound,yetsubtle,overlapbetweenprogramming,computersciencefundamentals,andCSS.ShespreadsherenthusiasmforthistopicatherdayjobasaDesignOpsEngineerforPenskeMediaCorporation,wheresheleadsthedesignsystemseffortforbigmediapublicationspoweredbyWordPress,andthroughspeakingandcommunityinvolvement.
Larawritesaboutherworkandlifeonherblog,illustratesfriendlymonsters,andinspiresallofuswhoworkwiththewebtoseekadeeperunderstandingofthecodewewriteandthetoolsweuse.
PerformanceMatters
9/142:30pm-3:10pm,StifelTheatreEncore
Performanceclearlymatterstousers.Forexample,themostcommonsoftwareupdateontheAppStoreis"Bugfixesandperformanceenhancements."NowthatMoore'sLawhasended,programmershavetoworkhardtogethighperformancefortheirapplications.Butwhyisperformancehardtodeliver?
Iwillfirstexplainwhycurrentapproachestoevaluatingandoptimizingperformancedon'twork,especiallyonmodernhardwareandformodernapplications.Ithenpresenttwosystemsthataddressthesechallenges.Stabilizerisatoolthatenablesstatisticallysoundperformanceevaluation,makingitpossibletounderstandtheimpactofoptimizationsandconcludethingslikethefactthatthe-O2and-O3optimizationlevelsareindistinguishablefromnoise(sadlytrue).
Sincecompileroptimizationshaverunoutofsteam,weneedbetterprofilingsupport,especiallyformodernconcurrent,multi-threadedapplications.Cozisanew"causalprofiler"thatletsprogrammersoptimizeforthroughputorlatency,andwhichpinpointsandaccuratelypredictstheimpactofoptimizations.Coz'sapproachunlockspreviouslyunknownoptimizationopportunities.GuidedbyCoz,weimprovedtheperformanceofMemcached(9%),SQLite(25%),andacceleratedsixotherapplicationsbyasmuchas68%;inmostcases,thisinvolvedmodifying<10linesofcodeandtookunderhalfanhour(withoutanypriorunderstandingoftheprograms!).CoznowshipsaspartofstandardLinuxdistros(aptinstallcoz-profiler).
EmeryBerger@emeryberger-gh:emeryberger-UniversityofMassachusettsAmherst
EmeryBergerisaProfessorintheCollegeofInformationandComputerSciencesattheUniversityofMassachusettsAmherst,theflagshipcampusoftheUMasssystem.HegraduatedwithaPh.D.inComputerSciencefromtheUniversityofTexasatAustinin2002.ProfessorBergerhasbeenaVisitingScientistatMicrosoftResearch(whereheiscurrentlyonsabbatical),theUniversityofWashington,andattheUniversitatPolitècnicadeCatalunya(UPC)/BarcelonaSupercomputingCenter(BSC).ProfessorBerger'sresearchspans
programminglanguages,runtimesystems,andoperatingsystems,withaparticularfocusonsystemsthattransparentlyimprovereliability,security,andperformance.HeandhiscollaboratorshavecreatedanumberofinfluentialsoftwaresystemsincludingHoard,afastandscalablememorymanagerthatacceleratesmultithreadedapplications(usedbycompaniesincludingBritishTelecom,Cisco,CréditSuisse,Reuters,RoyalBankofCanada,SAP,andTata,andonwhichtheMacOSXmemorymanagerisbased);DieHard,anerror-avoidingmemorymanagerthatdirectlyinfluencedthedesignoftheWindows7Fault-TolerantHeap;andDieHarder,asecurememorymanagerthatwasaninspirationforhardeningchangesmadetotheWindows8heap.HishonorsincludeaMicrosoftResearchFellowship,anNSFCAREERAward,aLillyTeachingFellowship,theDistinguishedArtifactAwardforPLDI2014,MostInfluentialPaperAwardsatOOPSLA,PLDI,andASPLOS,threeCACMResearchHighlights,aGoogleResearchAward,aMicrosoftSEIFAward,andBestPaperAwardsatFAST,OOPSLA,andSOSP;hewasnamedanACMDistinguishedMemberin2018.ProfessorBergeriscurrentlyservinghissecondtermasanelectedmemberoftheSIGPLANExecutiveCommittee;heservedforadecade(2007-2017)asAssociateEditoroftheACMTransactionsonProgrammingLanguagesandSystems,andwasProgramChairforPLDI2016.
RealtimeHybridReinforcementLearningatScale
9/142:30pm-3:10pm,StifelTheatreOvation
Nextbestaction(NBA)isatechniquethattakesuniqueuserhistoryandcharacteristicsintoconsiderationandrecommendsthenextactionsthathelpthecustomerprogressingtowardsbusinessgoalsasquicklyandsmoothlyaspossible.ItisnoteasytodesignsuchaAIpoweredNBAengine.Ideallyahand-freeNBAengineneedstohandlethefollowingproblems.a)Itshoulddealwithincompletehistoricalfeedbackthatareskewedtowardsasmallsetofactions;b)Itshouldadapttodynamicactions,whichcanbeaddedorremovedfrequentlyduetoseasonalchangesorshiftsinbusinessstrategies;c)Itneedstooptimizeformultiplecomplexbusinessobjectives,whichusuallyconsistofreachingasetoftargeteventsormovinguserstonextmorepreferredstage;d)Mostimportantly,ithastolearnandmakedecisionsinrealtimeandatmassivescale.Mostofthesolutionsinthemarketonlyaddressesafewofthosechallenges,duetomodelortechnicalchallenges.Inthispresentation,wewillshowhowweaddressallthoseissuesatSalesforceMarketingCloudEinstein.Wewillpresentahybridmodelbasedonreinforcementlearningthatbalancesbothonlineandofflinelearning.Wewillshowhowweutilizedistributedbigdataprocessingtechnologiesandservicestotrainandmakepredictionsatmassivescale.Wewillalsodiscussanofflineevaluationmechanismtoprovideboundedexpectedperformance,whichhasbeenahardproblemforreinforcementmachinelearningingeneral.
KexinXie@realstraw-gh:realstraw-Salesforce
KexinXieisaSeniorDirectorofEngineeringatSalesforce,responsiblefordatascienceresearch,practicesandarchitectureforMarketingCloudEinstein.HeleadstheteamforpushforwardtheAIinitiativeatverylargescale,thedatasystemprocessesover7billionmonthlyuniqueusersmakingtensoftrillionsofweeklypredictions.BeforeSalesforce,hehasworkedindifferentsectorsbuildinglarge-scaledatascienceplatformsthatperformmachinelearning,dataminingandonline/offlineanalyticsinthespaceofdatamanagement,real-timebidding,intelligentmarketing,anti-fraudandanti-moneylaundering.HehasaPhDincomputerscience,andhaspublicationsintoptierjournalsandconferenceslikeACMTODSandVLDB.
YuxiZhang-gh:juziyt-Salesforce
YuxiZhangisaSeniorDataScienceEngineeratSalesforceMarketingCloudEinstein.ShehasdesignedanddevelopedseveralAI-drivenfeaturesandrecommendationsystemsthatarecurrentlyservedtoMarketingCloudcustomers.PriortoSalesforce,shehasworkedonmusicrecommendationsystems,audiofingerprintingandA/BtestingatPandoraforafewyears.ShereceivedaMaster'sinMusicTechnologyandRobotics.
Zipline-ADeclarativeFeatureEngineeringLibrary
9/142:30pm-3:10pm,StifelTheatrePeabodyLounge
ZiplineisAirbnb'sdatamanagementplatformspecificallydesignedforMLusecases.Previously,MLpractitionersatAirbnbspentroughly60%oftheirtimeoncollectingandwritingtransformationsformachinelearningtasks.Ziplinereducesthistaskfrommonthstodays-bymakingtheprocessdeclarative.Itallowsdatascientiststoeasilydefinefeaturesinasimpleconfigurationlanguage.Theframeworkthenprovidesaccesstopoint-in-timecorrectfeatures-forboth-offlinemodeltrainingandonlineinference.
Inthistalkwewilldescribethearchitectureofoursystemandthealgorithmthatmakestheproblemofefficientpoint-in-timecorrectfeaturegeneration,tractable.
NikhilSimha-Airbnb
NikhilisaSoftwareEngineerontheMachineLearninginfrastructureteamatAirbnb.HeiscurrentlyworkingBighead,anend-to-endmachinelearningplatform.PriortoAirbnb,heworkedonschedulers,resourcemanagers,real-timedataprocessingenginesandcompilers.Heisalsotheco-authorofRealtimeDataProcessingatFacebook(SIGMOD-16).NikhilgothisBachelorsdegreeinComputerSciencefromIndianInstituteofTechnology,Bombay.Whilenotworking,helikestoboulderorplaycapoeira.
Machinelearningtointerpretsignlanguage
9/142:30pm-3:10pm,StifelTheatrePremier
InsteadofjustteachingDeafpeoplehowtogetalongwiththerestoftheworldandunderstandthem,whydon'twemakeanefforttounderstandthemandtheirlanguageviatechnology?!Inthistalkwewilllearnhowtousemachinelearningtointerpretsignlanguage.
TherearealargenumberofDeafpeopleacrosstheglobeandcommunicatingwiththemisdifficultasnoteveryonecanunderstandsignlanguage.InIndiatheofficialnumberofcertifiedsignlanguageinterpretersstoodatmerely250intheyear2017.Thistalkwillbeaboutusingobjectdetectionforagoodcauseandaddressingtheproblemofsignlanguagerecognition.Iwillalsoinvitecontributorsandencouragepeopletocontributeintheirownlanguagestomakethesystemmorerobustandaccessibleforpeopleworldwide.
ShaguftaGurmukhdas@shaguftamethwan
IamaTech-SpeakerforMozillaandhavebeenvolunteeringforMozillaforabout4years.IhavebeenexploringdeeplearningsincethepasttwoyearsandithasindeedbeenthefirsttimetechnologyhasmademefeelsoexcitedeversinceIcametoknowabouttheinternet.Otherthanthat,IamtheinitiatorandorganizerofDjangoGirlsPune.Iamalsoadecentartist,andlovetoplaythepianoinmyfreetime!
CommunityDrivenDevelopment
9/142:30pm-3:10pm,StifelTheatreTheater
Asdeveloperswhowanttolendourskillstosupportorganizationsfightingforsocialjustice,howdowebuildeffective,impactfulcollaborationswithorganizationsandmaketheresultingopensourceprojectssustainable?Overtwoyearsand1,062commits,we'llfollowtheevolutionofanopensourceprojectbuiltwithandforNewSanctuaryCoalition(NSC),anNYCimmigrantrightsorganization,tomeetexponentiallygrowingdemandfortheirimmigrationcourtaccompanimentprogram,proselegalclinic,andanti-detentionprogramfollowingthe2016election.
We'llexaminehowhavingtechnologistsembeddedinNSCprovidedastrongfoundationforcollaborationthroughfirst-handknowledgeoftheworkthesoftwareissupporting,motivationofworkingalongsideacommunity,andmutualtrustwithleadership.Reflectingonthewiderangeofskillsetsandexperiencelevelsbroughttotheproject,we'lllookathowthesuccessoftheprojectwasdependent,notjustoncodecontributions,butontrainingmaterials,laptopsetup,projectmanagement,digitalsecuritywork,andsetupofsupplementaryout-of-the-boxsoftware.DrawingfromNSC'sexpertiseinbuildingcommunityandlessonslearnedasanopensourcemaintainer,we'lldiscussthestrategiesweimplementedtobuildacommunityoftechnologiststosupportNSC'simmigrationjusticework.
ChristineZagrobelny-gh:CZagrobelny-FracturedAtlas
ChristineisaRubyengineer,opensourcemaintainer,andtheatredirector.AsanengineeratFracturedAtlas,shebuildstoolstohelpartistswiththebusinesssideoftheirart(fundraising,sellingtickets,findingstudio/performancespace).Outsideofwork,ChristinemaintainsopensourcesoftwaretosupporttheworkofanNYCimmigrantrightsorganization(NewSanctuaryCoalition)andisanorganizerforRailsBridgeNYC.Sheisdrivenandinspiredbythepotentialoftechnologytoenablesmallorganizationswithlimitedresourcestomaximizetheirimpact.
Jagged,ragged,awkwardarrays
9/142:30pm-3:10pm,UnionStationGrandF
Dataprocessinglanguages,suchasSQL,R,MATLAB,andNumpy/Pandas,implicitlyloopoveridentically-typedobjects("rows")ofadataset("table").Thismakesforasuccinctsyntaxinaninteractiveenvironment,butwhatdoyoudoifyourtabledoesn'thavearegularshape?
Particlephysicistshavethisproblem:eachcollisionofhigh-energyprotonscanproduceadifferentnumberofelectrons,photons,quarks,andotherparticlespecies.Asatablewithonecollisionperrow,thisdatasethasajaggededgeofunequal-sizedrows,sometimesreferredtoasaraggedarray.Traditionally,physicistshaveusedgeneral-purposeprogramminglanguageslikeFORTRANandC++todealwithbig,irregularlyshapeddatasets,butatalossofinteractivityandabstraction.
Tobringhigh-leveldataexpressivitytoparticlephysics,mygrouphasbeendevelopingawkward-array,alayeroverNumpythatgeneralizesitsarrayprogrammingparadigmtojaggedandotherawkwarddatastructures.AnyJSON-likedata,evenwithnested,heterogeneouscontent,canbesliced,broadcasted,andreducedwithimplicitloopsasthoughitwereaNumpyarray.
Thisgeneralizationofarrayprogramminghasimplicationsbeyondphysics:itsimplifiescombinatoricsandlikelihoodmaximizationsingenomicsandmayalsomakeiteasiertoanalyzestructuredlogfiles.MostawkwardarrayoperationscanbevectorizedtorunefficientlyonGPUs,andweareintegratingthelibrarywithApacheArrow,Parquet,Numba,andPandas.
JimPivarski-gh:jpivarski-PrincetonUniversity
JimwastrainedasaparticlephysicistwithaPh.D.fromCornellandhelpedcommissiontheCMSexperimentattheLargeHadron
Collider(LHC).HethenworkedasadatascientistforOpenDataGroupfor5years.In2016,hejoinedPrincetonasacomputationalphysicist,wherehedevelopedapopularsoftwarepackagelinkingparticlephysicsdataformatswiththescientificPythonecosystem,andisseekingnewwaystofostercommunicationandcodereusebetweenparticlephysicistsandotherfieldsofdataanalytics.
RhapsodyinZeroKnowledge:ProvingWithoutRevealing
9/142:30pm-3:10pm,UnionStationRegencyAB
Whatdoesitmeantoprovesomething?Propositionallogicrequireswerevealpropositionsandthenprovethemtrue,butwhatifitwerepossibletocreateproofswhichrevealnothingbuttheirownvalidity?Thisisthecoreideabehindtheemergentfieldofzeroknowledgeproofs:adecadesoldsolutiontoaproblemwedidn'tevenknowexistedyet.Asweincreasinglyturntomachinelearningtosolveproblemsrelatedtosecurity,suchasdetectingfrauds,fakes,orotherthreatactors,solutionsincreasinglyrelyonaccumulatingmassiveamountsofpersonaldata,underminingprivacyandriskingdatabreaches.Thistalkisabouthowwedon'thavetosacrificeprivacytogetsecurity,andthepotentialpowerofconstraints,complexity,anduniversalverifiablecomputation,andassumesnopriorknowledgeofcryptographyorproofsystems.
TonyArcieri@bascule-gh:tarcieri-iqlusion
Co-founder@iqlusioninc,formerly@[email protected],polyglotprogrammer,keymanagementwrangler,andinfrastructuresecurityspecialist
¡Escuinclababosa!APythonDeepLearningTelenovela
9/142:30pm-3:10pm,UnionStationRegencyC
Telenovelasarebelovedfortheiroverthetopdramaandintricateplottwists.Inthistalk,we'llreviewpopulartelenovelastosynthesizeatypicaltelenovelaarcanduseittotrainadeeplearningmodel.
Whatwouldatelenovelascriptlooklikeasimaginedbyaneuralnetwork?Toanswerthisquestion,we'llexaminethreePythondeeplearningframeworks-Keras,PyTorch,andTensorFlow-todeterminetheprocessoftranslatingatelenovelaintoaneuralnetworkandultimatelydeterminewhichonewillbebestforthetaskathand.Bepreparedforamor,pasiòn,andyelmisterioso!
LorenaMesa@loooorenanicole-gh:lorenanicole-GitHub
oliticalscientistturnedcoder,LorenaMesaisadataengineeronGitHub'ssoftwareintelligencesystemsteam,DirectoronthePythonSoftwareFoundation,andPyLadiesChicagoco-organizer.Lorena'stimeatObamaforAmericaandhersubsequentgraduateresearchrequiredhertolearnhowtotransformmessy,incompletedataintointelligibleanalysisontopicslikepredictingLatinxvoterbehavior.It'sthisuniquebackgroundinresearchandappliedmathematicsthatdroveLorenatopursueacareerinengineeringanddatascience.Onepartactivist,onepartStarWarsfanatic,andanotherpartTrekkie,Lorenaabidesbythemottoto"livelongandprosper".
InferenceQL:AIfordataengineers,withoutthemath
9/143:30pm-4:10pm,StifelTheatreEncore
CanweempowerdataengineerstouseAItoexplore,monitor,clean,andpredictdatastreams,withouthavingtolearnmath?
InferenceQLisanewopen-sourceAIplatformforsemi-structureddatathatisbeingdevelopedatMIT.InferenceQLuserscanbuildmodelsusingautomaticmodeldiscovery,thenquerythesemodelsusingasimple,SQL-likeAPI.InferenceQLalsoprovidesaspreadsheetinterfacewithbuilt-indatavisualization.InferenceQLrunsonbothJavaScriptandtheJVM,soitcandriveinteractivedataexperiencesinwebpagesandalsobepartofenterprisedatapipelines.ModelsinsideInferenceQLarerepresentedasprobabilisticprogramsinMetaprob,anewprobabilisticprogramminglanguageembeddedinClojure/Script.Expertscanthusreadandcustomizemodelsusingprobabilisticprogrammingtechniques.TheInferenceQLteamisworkingonanopen-sourceprototypesuitableforindustryandcivicuseindataengineering,analyticsconsulting,anddatajournalism,aswellasscientificdataanalysisandresearchinprobabilisticprogramming,causalmodeling,andprobabilisticexpertsystems.
InferenceQLwasmadebysomeofthesamepeoplebehindBayesDBandEmpiricalSystems,firstpresentedatStrangeLoopin2015.UnlikeBayesDB,InferenceQLisnotadatabase,butinsteadcanbeembeddedalongsidedatatables,datastreams,andclient-sidewebapplications.Wearejuststartingtoseekexternalbetatesters,contributors,andvolunteers.
UlrichSchaechtle@Schaechtle-MIT
UlrichSchaechtleisaresearchscientistatMIT.HeleadstheresearchengineeringeffortsaroundInferenceQL.UlrichholdsaPhDincomputersciencefromRoyalHolloway,UniversityofLondon,aswellasanMScincomputingfromImperialCollegeLondonandaBScinappliedcongitivesciencesfromtheUniversityofDuisburg-Essen.UlrichleadsanMITteamfortheDARPASynergisticDiscoveryandDesign(SD2)programs.HehaspublicationsinmajorconferencesandjournalsforprogramminglanguagesandAI.HecurrentlyworksonapplicationsofInferenceQLtodatajournalism,psychiatry,andsyntheticbiology.In2018,UlrichwasselectedbyDARPAasa
DARPARiser,oneof50earlycareerscientistsdevelopingbreakthroughtechnologies.
TweetMyWeddingDress
9/143:30pm-4:10pm,StifelTheatreOvation
Inspiredbyaloveofrainbowsandsillytechprojects,Idecidedtomakemyownfirstwearabletech-aweddingdressfulloflightsthatcouldbecontrolledbytweetsfrommyfriendsandfamily!
ThistalkwillgooverthetechIused(suitableforabsolutebeginners),themistakesImadeandthelessonsIlearnt.FromusingtheTwitterAPI,writingmyownAPI,codingforanArduinoandwiringthewholethinguptotheworstbit-THESEWING.Iaimtosharethemessagethatitisoktomakemistakes,wealldo!Ifyoukeeptryingyoucanmakesomethingmagical!
JoFranchetti@thisisjofrank-gh:thisisjofrank-Miss
JoistheEngineeringEngagementManageratTrainline.SheispassionateaboutWebVR,PWAsandgreatCSS.She'sgot7yearsexperienceasafrontenddeveloperandhasworkedinvariouspartsofthetechindustryfromstartups,agencies,charitiestolargeorganisations.Sheisalsoamentorandorganiseratcodebar.iowheresheisabletoactonherpassionnotonlyforteachinggooduseofthewebbutalsoforimprovingthediversityandinclusivityofthetechindustry.
Meander:DeclarativeExplorationsattheLimitsofFP
9/143:30pm-4:10pm,StifelTheatrePeabodyLounge
Asfunctionalprogrammerswelovetopaylipservicetodeclarativedatatransformation.Yetthisallegiancerarelyplaysoutinouractualcode.Datatransformingcombinators(map,filter,reduce,etc)offersignificantexpressivepowerovermanualloopsandmutation,butwhencombinedincomplexways,losetheirdeclarativepower.Lostinawebofnestedpipelines,theshapeofourdatabecomesobscured;ourcodebecomeshardtofollow.Wewindup,yetagain,needingtoplaycomputerinourheadstounderstandwhatourcodeisdoing.
MeanderisanexplorationintotrulydeclarativedatamanipulationcreatedasalibraryinClojure.Borrowingideasfromlogicprogrammingandtermrewriting,Meanderallowsdeclarativedescriptionsofarbitrarilycomplexdata;enablingyoutosearch,match,remember,join,andtransformanypartofyourdatadirectly.ThistalkwillshowyouhowtoleverageMeandertodeclarativelysolvereal-worlddatatransformationproblems,giveyouinsightintohowMeanderremainsbothperformantandexpressive,andfinallycastavisionforwhatamoredeclarativefutureforfunctionalprogrammingmayhold.
JimmyMiller@jimmyhmiller-gh:jimmyhmiller-Adzerk
JimmyisaSeniorSoftwareEngineeratAdzerkwhereheusesClojuretobuilddatatransformationandanalysissystems.Inhissparetimeheenjoysexploringprogramminglanguagetheory,readinganalyticphilosophy,andwritingcodethatnoonewilleveruse.
HowYourCompanyCanHelpSustainOpenSource
9/143:30pm-4:10pm,StifelTheatrePremier
Wouldyoulikeyourcompanyandco-workerstobegivingbacktotheopensourceprojectsyouuseeveryday?Inthistalk,Iwillpresentanumberofdifferentwaysthatcompaniesarecontributingtoopensource,anddiscusstheprosandconsofeach,sothatyoucanhelpyourcompanytosustaintheprojectsuponwhichyoudepend.Thiswillinclude:
Variousexcitingcorporateinitiativesforfundingopensourceprojects,andhowtochoosewhichprojectsgetthatmoneyWhatanOSPO(OpenSourceProgramOffice)isandwhyyourcompanymightwantoneWaystoencourageemployeestocontribute,andhowtomakecontributingeasyforthemHowtomeasureyourcompany'sandemployees'contributionstoopensource-Thisincludesphilosophicalquestionsofwhatshouldyoubemeasuring,andthelogisticalanswersofwhattoolsareavailabletodosoYouwillleavethistalkwithspecific,practicalideasandexamplesyoucanusetopromoteopensourcesustainability.
DanielleGellis@danisyellis-gh:danisyellis-Indeed
AstheToolsEngineerforIndeed'sOpenSourceProgram,Daniisbuildingtoolsthathelppeoplecontributetoopensourceprojects.She'spassionateaboutsharingandcollaboration,especiallyinsoftwareandtechnology.Daniisalsoateacherwhohastaughtbothkidsandadultstocode.
WatchtheWatcher:Facial-Recognition&PoliceOversight
9/143:30pm-4:10pm,UnionStationGrandF
Increasedstateuseoffacialrecognitiontechnologythreatenstoincreasethepowerdynamicbetweenthestateanddisenfranchisedcommunities.However,whathappenswhenindividualsusefacialrecognitiontowatchthewatchers?OpenOversightisanopen-sourceprojectthatpromotespoliceaccountabilitythroughpublicdata.Userscansortthroughphotostoidentifyanofficeragainstwhomtheywouldliketocomplain.Alackofofficeridentificationinformationhasstifledpreviousattemptstomonitorpolice;inChicagofrom2011-2015,28%ofcomplaintsagainstpolicewereimmediatelydroppedduetoalackofidentification.TheOpenOversightteamhasleveragedfacialrecognitiontomaketheprojectmoreeffective.Forexample,weusetheAWStoolRekognition,whichhasbeenusedbyvariouspolicedepartments,toidentifywith90%accuracyifpoliceormilitaryofficersarepresentinaphoto.Thispre-processinghasmadethevolunteerjobofphotosortingmoreefficient.Wealsouseface-api.jstomatchnewphotosagainstexistingimagesinthedatabase.Afteradiscussionofthistechnicalimplementation,thetalkwillclosebydiscussinglegalconsiderationsthatweencountered,suchasdepartment-specificguidelinesandcompliancewithIllinois'sBiometricInformationPrivacyAct.Thetalkwillprovideconcretetoolsandstrategiesforcommunityresponsestostatesurveillance.
EileenMcFarland-gh:McEileen
EileenisacurrentparticipantattheRecurseCenter,whereshefocuseshertimeonwebappdevelopmentandmachinelearningwithfeelings.Shehasworkedasasoftwaredeveloperonavarietyofprojects,usingJava/Spring,JavaScript,andAWS.Withabackgroundinlegalaid,Eileenremainsinvolvedincivictechprojects,suchasOpenOversight.Whennotprogramming,sheenjoysdoingyoga,reading,andpeople-watching.SheoncedroppedhercameraintoawaterfallinEcuadorafterrappellingdownaforementionedwaterfall.Thecamerastillworks,andEileenlikestoviewitasareminderoftheimportanceofresilience.
#langwishfulthinking
9/143:30pm-4:10pm,UnionStationRegencyAB
Whatmightsolvingarealproductionprobleminalanguagespecificallydesignedforbuildinglanguagesthatcooperatelooklike?RacketletsyouprogrambyfusingmultipleDSLstogether,reshapingandmouldingyourlanguagetofitthedomain,puttinguptowersoflanguages,infact,eachlevelclosertothelanguageofthedesireddiscourse.Getaglimpseofhowthismightworkbyimplementingsomethingthat'ssimpleyetspansawholerangeontheabstractionladderfrombittwiddlingtoWebframeworks.LearnhowtoreallygooffonatangentwhileimplementingFastCGIprotocolinRacket:wewon'ttieourselvestothedefaultsthatRacketdesignersblesseduswith,butboldlyemploywishfulthinking,borrowreadilyfromotherlanguages.Wewon'trushtomapdomainvocabularyontowhatbasicbuildingblocksourlanguageprovides-we'lltrytoactuallyspeakthelanguageofthedomain.WantprototypeswithLua-stylemetatables?Concisesyntax?Singleandmultipleinheritancewithgenericdispatch?Howgenericshouldthatgenericbe?Multimethods?FullMetaobjectProtocol,maybe?BeyondMetaobjectProtocol?Alanguageofpatternstoavoidbittwiddling?ThatoneWebframeworkthatspeaksthelanguageoftheWebyetletsyouwriteplain"desktopapp"code?Buildyourselfabetterlanguage.Daresay:IwishIcould-thenmakeitso.
VladKozin@zeRusski-gh:vkz-None
DilettanteprogrammerfromLondonwhotaughthimselfprogrammingwithRacketthroughHtDPandPLAI,didsomepaidJavascript,whichhedoesnotrecommend,thenpaidClojure,whichhedoes.HehasnowgonebacktotherootsandwriteslotsofRacket.Former@[email protected]'13@recursecenteraka@hackerschoolalum.
PatternMatching@ScaleUsingFiniteStateMachine
9/143:30pm-4:10pm,UnionStationRegencyC
Workingwithdataoftenmeanstryingtolocatedatathatfitspatterns,akintofindinganeedleinahaystack.Whenweaddbigdatafromnonhomogenoussourcestothemix,thisproblembecomesexponentiallycomplex.OneoftheusecasesatNetflix,isaboutimprovingtheSignUpexperiencethroughexperimentation.Beingabletofinduserjourneysacrossbillionsofevents;thatfollowcertainpatterns,isakeyinsightintosimplifyingthesignupprocess.
ThisgaveusanideatobuildaframeworktoexpresstheseuserjourneypatternsthatcouldbetranslatedintoaNonDeterministicFiniteStateMachine.OneoftheideasthatweadaptedfromKenThompson's1968CACMpaper,wastocreateanNDFAaroundpatternsdefinedusingregexthatcouldsupportbacktracking.ThenextstepwasapplyingthestatemachineacrossbillionsofeventsatscaleusingSpark.ThefinalpiecetothepuzzlewastomakeiteasilyusablebyDataEngineers,ScientistsandAnalystsalike.
Inthistalk,wewillcoverhowwebuiltthisframework(dubbed"Conduit")andthedesigndecisionsresultingfromchallengesalongtheway.Wewillalsotalkabouthowthiscanbeadaptedtorealtimeapplicationsinthefuture.
AjitKoti@ajitkoti-gh:ajitkoti-Netflix
AjitKotiisaSeniorEngineerontheGrowthDataEngineeringteamatNetflix,buildingproductsthatenableNonMemberAcquisition&Experimentation.Hehasover14yearsofexperiencebuildingandarchitectinglarge-scaledistributedsystemsandservices.AjithaspreviouslybuiltBigDataSolutionsforFanaticsandIBMLabs.
RashmiShamprasad@rshamprasad-gh:rshamprasad-Netflix
Passionateaboutallthingsdata,RashmiShamprasadisaSeniorDataEngineerontheGrowthDataEngineeringteamatNetflix,buildingdataproductsthatenableNonMemberAcquisition&Experimentation.Withover9yearsofexperienceworkinginBigData,herpreviousstintsincludebuildingBigDatasolutionsatPayPalandeBay.RashmiholdsaMastersinComputerApplicationsandBachelorsinCommerce.
ClosingKeynote
9/144:30pm-5:20pm,StifelTheatreTheater
Keynoteandperformance
ImogenHeap@imogenheap
LondonbasedrecordingartistÂImogenHeapblurstheboundariesbetweenpureartformandcreativeentrepreneurship.Writingandproducing4soloalbums,oneasÂFrouFrouÂ(withÂGuySigsworth),andcollaboratingwithÂJeffBeck,MikaandÂJoshGrobanÂamongstothers,Heaphaspennedtracksformovies,TVshowsandproducedthescoreforHarryPotterandtheCursedChild,winningthe'OutstandingMusicinaPlay'DramaDeskAward.

Counting5Grammynominations,winningoneforengineeringandanotherforhercontributiontoÂTaylorSwift's'1989',HeapalsoreceivedanIvorNovelloAward,TheArtistandManagerPioneeraward,theMPGInspirationAwardandanhonoraryDoctorateofTechnologyforherÂMI.MUgloveswork:aground-breakinggesturalmusicmakingsystem.

In2014sheenvisionedaflourishingmusicindustryecosystemthroughMyceliaandreleased'TinyHuman',thefirstsongtousesmartcontractsonablockchain.

Creatinganartist-led,fairandsustainabledecentralizedecosystem,Mycelia's'TheCreativePassport'providesanIDformusicmakerstoconnectdigitallywiththemusicindustry.Withthreeworldtours,soldoutRoyalAlbertHallandGreekTheatreshowsandthousandsof5*reviews,inSeptember2018Heapembarkedonayear-longmusicandtechnologyworldtour.