亚马逊游戏工作室如何在 AWS 上构建 AI_ML 数据基础.pdf

1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Tulip Gupta(She/Her)Principal Solutions Architect,AWS Strategic AccountsVuk Pejovic(He/Him)Enginee

2、ring Manager,Amazon Game StudiosHow Amazon Game Studios Builds a Data Foundation for AI/ML on AWSI N D 3 3 1 1 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AgendaUse case overviewData Forge solution overviewArchitecture whiteboard 2025,Amazon Web Services,Inc.or its affiliates.

3、All rights reserved.Petabyte-scale warehouse to process data such as player profileLifecycle management of ML models and jobsRapid onboarding and deployment of generative AI agents and assistants for fraud detection,scrubbing PIIUse case 2025,Amazon Web Services,Inc.or its affiliates.All rights rese

4、rved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Amazon Game Studios Data Forge story 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Code sharing and collaboration between teamsCI/CD pipeline managementCode discoverabilityCompatibility issuesGrowth challeng

5、es of data science workflow 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.SolutionData ForgeMono repo platformProvides an end-to-end infrastructure for model development,training,evaluationStreamlines the ML lifecycle 2025,Amazon Web Services,Inc.or its affiliates.All rights res

6、erved.80%Reduced model production time by D A T A F O R G E 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Data Forge platformPetabyte-scale warehouses(Redshift)with analytics layer on top,with Tableau as the frontendMLOps platform built on top of SageMaker internals that allows