Feature Stores and Evaluation Stores: Better Together
Josh Tobin from Gantry talks about the evaluation store and why to combine it with the feature store for more robust ML systems.
Josh Tobin from Gantry talks about the evaluation store and why to combine it with the feature store for more robust ML systems.
This event focuses on the role that feature stores play in putting ML models into production, but feature stores alone are not enough to build production machine learning systems.
In this talk, we will define the *evaluation store*, a category of machine learning infrastructure that serves as the single source of truth for model evaluation metrics across all phases of the model lifecycle. Then we will discuss the connection between feature stores and eval stores, and why we believe that these two tools must work together to create robust machine learning systems.