scikit-learn ML Pipelines

Build reproducible ML workflows with scikit-learn Pipelines, ColumnTransformers, cross-validation, and MLflow experiment tracking.

Principles

• Prefer Pipeline/ColumnTransformer so preprocessing travels with the model
• Make runs deterministic: set random_state everywhere and seed numpy
• Keep train/val/test separation; use cross-validation for small datasets
• Persist the whole pipeline with joblib and load it for inference

Project Layout