How Rescale built an AI triage agent with Strands Agents and AI-DLC
aws ·
Rescale, a digital engineering platform for the AI era, developed an intelligent triage agent to automate the classification and routing of nightly end-to-end test failures. Utilizing the open-source AI-Driven Development Lifecycle (AI-DLC) framework, the team built the solution in just five days from problem definition to staging deployment. The agent leverages Claude Opus 4.8 on Amazon Bedrock to categorize root causes such as infrastructure issues or code bugs, while retrieving historical patterns and pod logs via Model Context Protocol. Built with the Strands SDK and hosted on Amazon EKS, the system integrates directly into Rescale’s existing Argo Workflows CI/CD pipeline. This automation has increased classification agreement with human triage from 44% to 68% and reduced production blocking time by 20%. By adopting a mob-building approach where humans validate AI proposals in real-time, Rescale ensured architectural rigor without slowing development. The success of this initiative has led Rescale to establish an internal Center of Excellence to scale AI-DLC practices across its broader engineering organization.
aws