Building Agentic AI Systems with AWS Serverless
summary
Shows a concrete pattern: wrap company APIs as tools, give the LLM context and run model→tool→model loops to complete goals — demonstrated with a travel agent that starts with flight booking and scales to car/hotel by extending tools. Explains tradeoffs: agent decides tools and stop (non‑deterministic) versus deterministic AI workflows, and recommends Lambda, Step Functions, Amazon Bedrock (plus S3/DynamoDB, API Gateway, CloudWatch, KMS) or an open SDK runtime. Stresses observability via Step Functions loops and the need to verify with LLM-as-judge before production.
AI-written summary · about
chapters (8)
- 0:00Intro
- 1:10Agenda & takeaways
- 2:07What are AI agents?
- 3:29How are AI agents different from AI workflows?
- 11:50Building AI agents on AWS
- 21:16Things to consider
- 22:16Resources
- 22:57Outro
related talks
28mBuilding Agentic AI applications on AWS
45mServerless & Event-driven Patterns for GenAI
48mHow to Make Smart Architecture Decisions when Building Gen AI Apps
23mBuilding Production Multi-Agent RAG Systems on Serverless AWS
51mTaking Serverless to the Next Level
31mAgentic AI architecture patterns
17mTransforming Robotics Service with Agentic AI on
41mServerless LLM with AWS Lambda and LangChain: A Revolution in Application Development