scikit-learn Tabular ML AI Skill Guide

Overview & Engine Architecture

scikit-learn standardizes estimators with fit / predict / transform. Pipeline and ColumnTransformer bundle preprocessing with models so cross-validation applies the same folds to transforms. Agents prevent leakage by fitting preprocessors inside CV, choose metrics matching the task, and persist entire pipelines - not only the final classifier.

raw columns -> ColumnTransformer
                    -> Pipeline(estimator)
                    -> cross_val / GridSearchCV
                    -> joblib dump

When to use this skill

• Classical ML on tabular features
• Baselines before deep learning
• Feature preprocessing that must ship with the model