atlas-llm-evaluation-engineering-build-llm-judge

Design, calibrate, and bias-audit an LLM-as-judge so its scores track human judgment: named criteria with anchored levels (not 'rate 1-10'), a structured verdict + reason, a pinned judge model/version, a human-agreement calibration on a sample, and checks for position / verbosity / self-preference / leniency bias. Reach for it whenever a judge grades outputs. Used by eval-harness-engineer (primary).