Model approval workflows in MLOps - Practice Problems & Coding Challenges

Challenge - 5 Problems

🔀 Workflow

intermediate

Identify the correct sequence in a model approval workflow

Which of the following sequences correctly represents the typical steps in a model approval workflow?

A 1,2,3,4

B 2,1,3,4

C 1,3,2,4

D 3,1,2,4

💡 Hint

Think about the logical order from building to deploying a model.

The model is first trained and evaluated, then validated by experts, followed by an approval decision, and finally deployed.

🧠 Conceptual

intermediate

Purpose of model approval in MLOps

What is the main purpose of having a model approval step in an MLOps pipeline?

A To reduce the size of the model for faster inference

B To speed up the deployment process by skipping testing

C To ensure the model meets quality and compliance standards before deployment

D To automatically retrain the model without human intervention

💡 Hint

Think about why we want to check a model before using it in real life.

Model approval ensures the model is reliable, safe, and meets required standards before it is used in production.

❓ Troubleshoot

advanced

Diagnose failure in automated model approval pipeline

An automated model approval pipeline fails to trigger the approval step after model validation. Which is the most likely cause?

A The validation results were not saved in the expected storage location

B The model training step completed successfully

C The deployment environment is offline

D The model size exceeds the maximum allowed limit

💡 Hint

Consider what the approval step depends on to start.

If validation results are missing or misplaced, the approval step cannot proceed because it lacks required input.

💻 Command Output

advanced

Output of model approval status check command

What is the output of the command `mlflow models get-approval-status --model-id 123` if the model is approved?

MLOps

mlflow models get-approval-status --model-id 123

A {"model_id": 123, "status": "pending", "approved_by": null}

B {"model_id": 123, "status": "approved", "approved_by": "team_lead"}

C Error: Model ID 123 not found

D {"model_id": 123, "status": "rejected", "approved_by": "qa_team"}

💡 Hint

Look for the status that indicates approval.

The output shows the model ID, status as approved, and who approved it.

✅ Best Practice

expert

Best practice for integrating human approval in CI/CD for models

Which approach best integrates a human approval step into an automated CI/CD pipeline for model deployment?

A Approve models based on training accuracy alone without human review

B Automatically deploy all models without any approval to speed up delivery

C Skip validation and rely on monitoring after deployment to catch issues

D Use a manual approval gate in the pipeline that pauses deployment until a designated person approves

💡 Hint

Think about how to combine automation with necessary human checks.

A manual approval gate pauses the pipeline and requires explicit human confirmation before continuing, ensuring safety and compliance.