World Model Platform

A compact Kubernetes-native platform for one model lifecycle:

collect → validate → train → evaluate → register → serve → observe → improve

The workload is a world model. The systems questions are general. Source and architecture live in the AI Infrastructure repo.

Notes coming.