Packages a trained model for deployment with a versioned artifact, an input/output schema, a smoke test, a deterministic preprocessing pipeline, and a deployment manifest. Triggers when a notebook-trained model is about to be served behind an API, batch job, or scheduled inference; when a teammate asks how to ship a model; when the only artifact is a bare pickle in a notebook directory; or when the user is choosing between joblib, ONNX, TorchScript, or a framework-native serialization format. Refuses to certify a model as deploy-ready without a captured input/output schema and a smoke test that round-trips through the saved artifact.
