AI Evaluation Engineer
How to build the project
Build twenty gold cases and a rater guide
0/3 build steps lit
If two humans disagree, fix the guide before you score the model.
20 cases · One-page rater guide · One caught lie
Week 1 · Twenty cases
Five of each kind.
Ship the artefact
Write what you built in one screen. Ship puts a verb-object-number line on your resume.