Module assessment

1.

A clinic's scheduling system needs an immediate no-show risk score the moment staff book an appointment. Which endpoint type should the team deploy the model to?

Batch endpoint

Online endpoint

Private endpoint

2.

Why can Azure Machine Learning generate the scoring script and environment automatically when a data science team logs a model with MLflow?

Because the MLflow model format stores metadata that Azure Machine Learning uses to build them during deployment

Because registering the model converts it to a pickle file that requires no additional configuration

Because Azure Machine Learning only supports deploying MLflow models

3.

A team shifts 25% of endpoint traffic to a new model deployment. Its error rate rises sharply, while the previous deployment remains healthy. What should the team do first?

Delete the previous deployment and recreate the endpoint

Route all traffic back to the previous deployment, then investigate the new version

Continue increasing traffic to determine whether the error rate stabilizes

4.

A team deploys and tests a model automatically, but wants a reviewer to approve it before the workflow shifts production traffic. What should the team configure?

A protected GitHub production environment with a required reviewer

A branch protection rule that runs after every deployment

A repository variable that records whether the deployment was approved

5.

A model monitor reports a sustained prediction drift signal, and clinic staff confirm the flagged patients aren't actually high-risk. What's the most appropriate next step for the data science team?

Ignore the signal, since the endpoint's operational metrics show a normal error rate

Investigate the signal and consider retraining the model on more recent data

Delete the endpoint so it stops serving predictions

Feedback