Feast is an open-source feature store which whylogs can easily integrate with. Feature stores are used for storing and managing a transformed version of data which is consumable by machine learning models. Learn more about feature stores here.
When monitoring data which feeds a machine learning model, users are recommended to monitor both the raw form of the data as well as the transformed version of the data which often lives in a feature store. Data quality degradation can be caused by both changes to the raw data or issues which occur during data transformations. By monitoring both raw and transformed data, users can more quickly diagnose the root cause of issues that result in poor model performance.
Users can visit Feast’s quickstart example to get up and running with Feast.
Integrating with whylogs with a Feast feature store is detailed in this example notebook from our whylogs repository.