MLOps Engineer interview question
What is a model registry and why do you need one?
A model registry stores immutable model versions with metrics, lineage, environment, approvals and lifecycle stage. It gives deployment a reviewed artifact identity and supports promotion, comparison, audit and rollback.
Direct answer
What a strong answer covers
How it works
A registry is the system of record for models the way a package repository is for libraries. Each entry is an immutable version with the artifact, its hash, the training run that produced it with code and data versions, evaluation metrics, the environment it needs, and a lifecycle stage such as staging or production with who approved the transition and when. Deployment pulls by version from the registry rather than from a file someone uploaded, so what runs in production is identifiable, comparable to its predecessor, auditable, and rollable back to. Without one, teams end up with model_final_v3_really.pkl in a bucket.
Example
When an audit asked which model produced credit decisions in March, the registry showed version 17, its training run with the data snapshot and commit, its approval by the risk lead with a linked evaluation report, and its promotion date. Rolling back a bad release in another team took one command because the previous version was still registered with its full environment.
Common mistakes
- Deploying from an artifact path with no link to how it was produced.
- Overwriting a registered version instead of creating a new one.
- Promoting without an approval record so no one knows who accepted the evaluation.
A useful answer structure
Answer from your own work.
Do not memorise a script: use only examples you can explain honestly in follow-up questions.
Open the desktop workspace and rehearse a clear answer in your own words.