Hopsworks

Hopsworks is an open-source feature store and MLOps platform for managing the full machine learning data pipeline — feature engineering, feature storage/versioning, model training, and model serving in one platform.

Its feature store specifically solves the training/serving skew problem (features computed differently for training versus live inference) by ensuring both paths use the exact same feature definitions and computation logic.

Used by ML teams running many models in production who need consistent, reusable, versioned features across projects rather than each model pipeline recomputing features independently.