This charter adheres to the conventions, roles and organization management outlined in wg-governance.
SIG Feature Store covers the definition, management, storage, discovery, and serving of features to models.
This SIG aims to coordinate projects and technologies necessary to enable the core functionality required to deploy and operate a feature store in Kubeflow.
- Ensure that users have a registry to define and manage features and their related metadata.
- Ensure that users have a means of data ingestion, management, and storage for the purposes of model training and online serving.
- Ensure that users have a unified feature serving layer for both model training and online serving.
- Ensure that users have the ability to both generate and validate feature statistics.
- Ensure that users have operational instrumentation necessary to safely run a feature store in production.
- Ensure that users have the documentation and tutorials necessary to both deploy, operate, and use a feature store.
- Ensure that Kubeflow maintains a cohesive data tooling vision with respect to feature stores.
- Coordinating with Pipelines/KFData WG to ensure both datasets and streams can be ingested, persisted, and served.
- Coordinating with Training WG to make sure that its possible to create training datasets using the feature store.
- Coordinating with Serving WG to make sure that its possible to retrieve online feature data from the feature store.
- Coordinating with Manifests WG to ensure that feature store manifests are properly deployed with Kubeflow.
- Coordinating with release teams to ensure that the feature store functionality can be released properly.
- Data pipelining, immutability, and lineage.
This SIG follows adheres to the Roles and Organization Management outlined in wg-governance and opts-in to updates and modifications to wg-governance.
SIG Technical Leads.