DeepBrew 2.0 星巴克机器学习平台.pdf

1、2024 Databricks Inc.All rights reserved1Vikas VennavaliVikas VennavaliJun 13,2024Jun 13,2024DEEPBREW 2.0STARBUCKS ML PLATFORM2024 Databricks Inc.All rights reserved2024 Databricks Inc.All rights reserved2Table of ContentsTable of ContentsIntroductionPrincipal ObjectivesPlatform ArchitectureML Pipeli

2、nes Model|Data|Function as a ServiceLogging,Monitoring&AlertingEnvironment SetupAncillary HighlightsCurrent&Upcoming InitiativesRecap&ConclusionQ&A2024 Databricks Inc.All rights reservedDeepBrew is a Starbucks home AI/ML platform that empowers scientists and engineers to deploy ML applications secur

3、ely and at scale with Databricks as primary compute for training3IntroductionIntroductionDEEPBREWDEEPBREW2024 Databricks Inc.All rights reservedPrincipal ObjectivesQuicker to innovation,shorten time to production insightsHigher quality products and services at scaleFoster better collaboration across

4、 different orgsEasy to share reusable codes,framework,and utilitiesDiscoverability of applications,models,and featuresReduce Data Scientist ops support significantly2024 Databricks Inc.All rights reserved2024 Databricks Inc.All rights reserved5Platform ArchitecturePlatform ArchitectureOverviewOvervi

5、ew5Analytics Playground WorkspaceStarbucks ApplicationsStarbucks Store Compute PipelinesDeepBrewTrainEvaluateBatch PredictRegisterDeployModel as a ServiceFunction as ServiceData as a ServiceMicroservicesArchitecture ManageAnalysis&ReportingPredictions2024 Databricks Inc.All rights reserved2024 Datab

6、ricks Inc.All rights reservedPlatform ArchitecturePlatform Architecture6ML PipelinesML Pipelines6OrchestratorADLSEvent HubBlob StorageRedisCosmos DBSQL Server2024 Databricks Inc.All rights reserved2024 Databricks Inc.All rights reservedPlatform ArchitecturePlatform Architecture7M