The current ML CI/CD pipeline only kicks off model updates on the hub cluster. We should prepare PRs for factory deployments.