Selective Reading Guide

Full-book reading is not required. Start with the guide, then use sources to resolve a named gap.

Need	Source	Read for
Pipeline and evaluation API behavior	scikit-learn User Guide	Exact estimator, split, calibration, and metric semantics
Data lineage and system boundaries	Designing Data-Intensive Applications	Derived-data reliability and reproducibility
Risk framing	NIST AI RMF	Govern, map, measure, manage vocabulary
Dataset documentation	Datasheets for Datasets	Motivation, composition, collection, use, maintenance
Model reporting	Model Cards	Intended use, evaluation groups, limitations

Record the question that triggered reading and the decision changed by it. A bibliography without a decision trace is not evidence.