Take a tabular business problem from raw data to a LIVE, reproducible /predict endpoint. Full brief: docs/curriculum/ds-deployed-e2e-mlops-brief.md. Video shell: courses/video-scripts/mlops-cloud-deploy/05-deployed-e2e-mlops-brief.md.
Pass bar
80+/100 overall AND reproducibility gate M6 >= 10/15. Reviewer cold-clones the repo and reproduces metrics.json within +/-0.005 plus a live /health check.
Do not start
Until the data-scientist P4/P5 milestone placeholder is flipped to required (owner action — this seed never flips it).