Build evaluation suites for LLM applications: golden datasets, metrics, LLM-as-judge, regression gates, and CI integration. Use for trustworthy model behavior.
