AWS Engineers Demonstrate AI Agent Control and Deployment
Elizabeth Fuentes Leone and Sandhya Subramani demonstrate building a customer service agent using Strands Agents, focusing on implementing hooks to limit tool calls, adding contextual skills, and deploying the workflow to Amazon Bedrock AgentCore.
ImportanceLocalEvidenceE2 unreplicatedWrite-upQuick
In the "Agent Speedrun" workshop, Elizabeth Fuentes Leone and Sandhya Subramani demonstrate how to build and control a customer service AI agent, starting from system prompts and basic tools. Published on October 11, 2026, by the ai.engineer channel, the video addresses the engineering challenges of agent orchestration.
The core content focuses on controlling agent behavior in production. Using the Strands Agents framework, the presenters show how to introduce hooks to limit tool invocation rates and verify these constraints through live testing. They also compare programmatic checks with instructions left to probabilistic models and explain how Markdown files provide contextual skills, such as order tracking procedures.
The latter half covers deployment and observability. The speakers migrate code from local notebooks to Amazon Bedrock AgentCore, detailing runtime setup, infrastructure generation, and session tracing configuration. This walkthrough illustrates the path from a small prototype to a monitored production service.