AI agents are dynamic systems shaped by prompts, models, tools, and runtime decisions. Operating them in production requires observability, tight feedback loops, and safe rollout mechanisms.
In this webinar, we’ll build a Strands-based agent, deploy it on Amazon Bedrock AgentCore, and instrument it with Datadog Agent Observability to trace prompts, tool calls, retrieval, and multi-agent interactions. We’ll then add evaluations to measure quality and feature flags to control behavior at runtime, enabling safe experimentation, gradual rollouts, and fast rollback.
Leave with practical patterns and a runnable reference for building and optimizing agents in production.
Senior Technical Advocate, Datadog
Principal Partner Solutions Architect, Amazon Web Services