Machine Learning Feature Store for Patient Predictions

Architect a machine learning feature store using Apache Cassandra and Node.js that can efficiently store and retrieve patient health features for predictive modeling. Design a schema that supports time-series health metrics, allowing efficient feature extraction for ML algorithms. Implement automatic feature versioning and lineage tracking, with support for A/B testing different feature sets. Create a real-time feature computation pipeline that can generate ML-ready datasets with minimal latency.