Build and scale reliable multi-agent systems on AWS
summary
Built a production multi-agent system using Amazon Bedrock foundation models with a Bedrock Agents multi‑agent pattern, a web search tool hitting the ATO site, a code interpreter for charts/files and a lightweight REST API wrapper, scaling by adding ~100 functions in a few sprints. The SaaS handles 300,000 funds and A$500B in assets, faced prompts bloating to 30 pages, and claims to reduce 2–3 hour tasks to 13 minutes by splitting responsibilities into task‑specific agents.
AI-written summary · about
chapters (10)
- 0:00Introduction and setup
- 1:26What are GenAI agents
- 3:34How agents work and plan
- 7:24Amazon Bedrock agents
- 11:03BGL’s multi-agent journey
- 16:41Single agent demo and benefits
- 21:19Challenges with single agents
- 22:23Multi-agent collaboration
- 25:18Long-running autonomous agents
- 27:45Summary and key takeaways
related talks
43mBeyond Chatbots: How Amazon Bedrock Agents Drive Enterprise AI
23mBuilding Production Multi-Agent RAG Systems on Serverless AWS
40mBuilding an AWS solutions architect agent with Amazon Bedrock
33mShipping Enterprise AI Agents with Amazon Bedrock AgentCore
44mAgentic AI at Sage with AgentCore
17mTransforming Robotics Service with Agentic AI on
19mBuilding Agents at Cloud Scale
01h 00mBuilding a Polyglot Multi-Agent AI Architecture