How to Know If Your AI Agent Is Any Good: Evals With Phoenix & Claude — Laurie Voss, Arize AI

Why watch

Laurie Voss demonstrates how to build a production-ready evals loop for Claude-powered AI agents using Arize Phoenix. The workshop covers tracing agent executions, writing code evals and LLM-as-a-judge evaluations, testing judge reliability, validating prompt fixes with experiments, upgrading AI models with evidence, and integrating evals into CI using npm test. Ideal for engineers building production AI agents who need systematic ways to catch confident but wrong outputs.