Model registry
A central record of trained model versions with their training data, parameters, evaluation results, approval status and deployment history. It is the system of record for which model is running where, and why.
Why it matters
Without a registry, it is hard to answer basic governance questions: which version produced a decision, what it was tested on, who approved it and how to roll back. Those questions come up in incidents, audits and complaints.
A registry also supports controlled release: models move through stages such as candidate, approved and production, with checks at each step.
In practice
For example, a UK insurer’s pricing team might register every model version with its validation report and approver, so when a pricing complaint arrives it can identify exactly which model and data version produced the quote.
Where Rodan fits
Rodan sets up model versioning, approval and deployment tracking as part of Platform and Cloud Engineering for AI systems.