Machine Learning Feature Store for Predictive Healthcare

Architect a scalable Python-based feature store for medical machine learning models using PostgreSQL and MLflow. Develop a system that can version, store, and retrieve complex medical feature sets while maintaining data lineage and enabling reproducible research. Implement a dynamic schema that supports heterogeneous medical data types, including genetic markers, clinical observations, and patient history. Include automated feature validation, drift detection, and a versioning mechanism that allows researchers to track feature engineering experiments.