建立对人工智能代理的信任：大规模自动化评估 [重复].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.Mohammad TahsinSpecialist SA,GenAIDaniel StairSpecialist SA,AnalyticsBuilding Trust in AI Agents:A

2、utomated Evaluation at ScaleO P N 3 0 4-R 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AgendaUse caseOverview of technologiesDemoQ&A 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Use case 20

3、25,Amazon Web Services,Inc.or its affiliates.All rights reserved.Use case Financial Research Agent Build a Financial Research Agent Agent has access to Tools RAG 10k Web Search Evaluate 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.

4、All rights reserved.Technology overview 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Technology overviewHayStackMLFlowSageMaker Pipelines 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Flexible pipelinesExtensible modulesFeatures of HaystackProduction ready

5、reliability 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Experiment trackingSageMaker MLFlowModel management&evaluationCollaboration across teamsProduction readiness 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Centrally manage each step of the workflowCho

6、ose from built-in templatesShare and re-run workflowsAmazon SageMaker Pipelines 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Amazon SageMaker Pipelines 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Amazon Bedrock has everything you need