ML deployment governance

The controls governing how model versions reach production, including approval gates, lineage records, immutable serving infrastructure and evaluation independent of the team making the change.