释放智能体人工智能的力量：借助 NVIDIA 和 AWS 推动任务成功（由 NVIDIA 赞助）.pdf

《释放智能体人工智能的力量：借助 NVIDIA 和 AWS 推动任务成功（由 NVIDIA 赞助）.pdf》由会员分享，可在线阅读，更多相关《释放智能体人工智能的力量：借助 NVIDIA 和 AWS 推动任务成功（由 NVIDIA 赞助）.pdf（22页珍藏版）》请在三个皮匠报告上搜索。

1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Tapping into the Power of Agentic AI:Driving Mission Success with NVIDIA and AWSChris HollandSr.Solutions ArchitectNVIDIA FederalAIM278-S 2025,Amazon Web Services,In

2、c.or its affiliates.All rights reserved.We solve customers biggest challenges with large scale,cost effective,and flexible GPU-accelerated solutions.PARTNERING ACROSS INFRASTRUCTURE,SOFTWARE,AND SERVICES&2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Evolution of Generative AILLM

3、sRAGAgents202420232025 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.NVIDIA Terms NeMO=Neural Modules for AI/ML pipelines NeMo Agent Toolkit=Orchestrate,measure and optimize agents NIM=NVIDIA Inference Microservices.Containers.Blueprints=Kubernetes Helm Charts of NIMs+Other AI S

4、oftware Goal=Run at scale in production 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.How AI Agents WorkAI Agents use Tools and Collaborate to Complete Complex ProjectsTHINKINGTOOL USECOMPUTER USEMEMORYI/O PROMPTOTHER AGENTS 2025,Amazon Web Services,Inc.or its affiliates.All rig

5、hts reserved.System-level accelerations require knowledge of the entire systemMany tools existbut dont work togetherRepeatabilityChallenging to guarantee consistent resultsBuilding Agentic Systems is ComplexAgentic applications should integrate with existing tools,heterogeneous data,and perform reli

6、ablyArchitectural ComplexityPerformanceCode ReuseFragmented solutions result in duplicate work 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.The prototype to production“chasm”Excitement and potentialMeaningful business valueChallenges on the path to productionPerformanceSecurity