The Open Data Hub community has a need for a Model Registry solution as part of our community-driven offerings; we are actively looking into potential solutions in order to offer a cloud-native Model Registry to be integrated as part of the overall ODH architecture. We want to leverage ML Metadata for this endeavor. We would like to collaborate closely with this community to not only improve MLMD but also to share any enhancements and the resulting Model Registry solution with this community for mutual benefit.

We would like to gauge the interest of this community in such a collaboration.
As part of this we have few basic questions such as:

• Is there a roadmap/backlog defined for the ML Metadata?
• Importance of local server deployment model where remote server deployment model is not always being used?
• Is there a working group, committee, or development team where we can bring up the issues/enhancements about ML Metadata?
• ...and others

Note: we understand that ML Metadata is part of TFX, and as part of TFX a “TensorFlow Hub” is available, but the requirements above for a Model Registry solution relate to a kube-native and deployable artifact, for a diverse types of Machine Learning models beyond specifically TF models. i.e.: a Model Registry deployable in any K8s cluster.

cc/ @dhirajsb @rareddy