【新品发布】SageMaker & MLflow：无需基础设施管理加速创新.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.AIM3340SageMaker AI&MLflow:Innovate faster without infrastructure managementRahul EaswarSenior Pro

2、duct Manager Amazon Web ServicesRohan TangadpalliwarStaff Software EngineerIntuit 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AI projects dont fail because of poor models they fail because of poor operations and missing governance.2025,Amazon Web Services,Inc.or its affiliates

3、.All rights reserved.AI/ML Lifecycle:The Role of AI GovernanceOnboardPrepare data assets and pipelines,Manage data quality and biasBuild&TrainExperiment and automate the execution of build pipelinesRegisterEvaluate,register,and approve ML models for deploymentDeployAutomate the execution of deploy p

4、ipelines into productionMonitorMonitor model performance post-deploymentCLASSIC MLGENAI MODELSOnboardPrepare data assets and pipelines,Manage data quality and biasPre-train or Fine-tunePre-train FM or Select from FM hub to fine-tuneRegisterEvaluate,register,and approve GenAI models for deploymentDep

5、loyAutomate the execution of deploy pipelines into productionMonitorMonitor model performance post-deploymentMANAGING LIFECYCLE WITH AI GOVERNANCE 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Key challenges addressed by AI governancePurpose Built Across the Entire AI/ML lifecyc

6、leNon-standardized infrastructure provisioning slows onboarding and risk security complianceCollaborate while controlling access to sensitive AI assetsEnsure models are explainable,transparent,and reproducibleOnboardDeploy/monitorBuildTrainPrepareTune 2025,Amazon Web Services,Inc.or its affiliates.A