Technical interview guide

Model Versioning & Registries

Tracking every trained model artifact alongside the data and code that produced it, so results are reproducible.

Read: 45 min
Practice MCQs: 25
Interview QA: 25
Edition: v4
Editorial status: Reviewed
Relevant for: MLOps Engineer

Scope: MLflow, SageMaker AI, Vertex AI, Azure Machine Learning, and NIST AI RMF guidance current 2026-09-01.

Interview QA

Treat each question like a live interview question: answer out loud first (structure, assumptions, tradeoffs), then open the model answer to spot gaps and rehearse a tighter follow-up.

Curated: 2026-10-05 · Written: 2026-10-05 · Reviewed: 2026-10-03

QA-1

What is a model registry, and why is standard code version control (such as Git) insufficient for versioning machine learning models in production?

QA-2

Define what constitutes one model version.

QA-3

Build end-to-end lineage for a registered model.

QA-4

Design model promotion and approval gates.

QA-5

Use model aliases safely in deployment workflows.

QA-6

Design a safe registry-backed rollback.

QA-7

Secure a model registry against supply-chain compromise.

QA-8

How do model aliases and tags enable zero-downtime progressive delivery compared to legacy stage transitions in modern model registries?

QA-9

Design registry organization across dev, staging, and production.

QA-10

Register and govern an ensemble model.

QA-11

Version a fine-tuned foundation model and adapters.

QA-12

Handle a registry alias that disagrees with the production endpoint.

QA-13

Design registry metadata schemas and quality controls.

QA-14

Plan cross-account or cross-region model sharing.

QA-15

Design registry support for online and batch deployments.

QA-16

How do you implement cryptographic signing and verification for model artifacts stored in a registry?

QA-17

Choose between promoting artifacts and promoting training code.

QA-18

Respond to a tampered or unsigned model artifact.

QA-19

Manage model deprecation, retention, and deletion.

QA-20

Design model cards and intended-use evidence in a registry.

QA-21

Handle non-deterministic training in registry reproducibility.

QA-22

Version models that depend on an online feature store.

QA-23

Design registry governance for emergency model promotion.

QA-24

Measure whether a model registry is improving MLOps outcomes.

QA-25