Skip to content

Model Deployment Questions

Practice questions for Model Deployment topic in Databricks Certified Machine Learning Associate. 24 questions covering this domain.

24 questions8 easy12 medium4 hard
Q1
easy

In the Unity Catalog Model Registry, what replaces the stage concept (Staging, Production, Archived) that exists in the Workspace Model Registry?

Q2
medium

To configure MLflow to use Unity Catalog as the model registry backend instead of the Workspace Model Registry, which configuration call is required?

Q3
easy

A data scientist has registered a model in the Workspace Model Registry and wants to expose it as a REST API for real-time inference. Which Databricks...

Q4
hard

A data science team deploys a model for batch inference using `mlflow.pyfunc.spark_udf()` on a DataFrame with 1 million rows. The job is very slow des...

Q5
medium

A data science team wants to automatically trigger a CI/CD job when a model version is transitioned to Production in the Workspace Model Registry. Whi...

Q6
medium

A junior data scientist requests a stage transition to Production in the Workspace Model Registry but does not have the required permissions. What hap...

Q7
hard

A data scientist registers a model in the Unity Catalog Model Registry and assigns the alias `champion` to version 5. Later, version 6 is trained and ...

Q8
easy

Which URI scheme is used to reference a model version by alias in the Unity Catalog Model Registry?

Q9
easy

What does a Mosaic AI Model Serving endpoint return when called with a POST request?

Q10
medium

A data scientist creates a new challenger model and wants to test it alongside the existing production model in Mosaic AI Model Serving without changi...

Q11
hard

A data science team uses Mosaic AI Model Serving to deploy a custom PythonModel. During load testing, the endpoint returns HTTP 429 errors during peak...

Q12
medium

A data scientist wants to register a model in the Unity Catalog Model Registry using the MLflow Python API after training. Which combination of config...

Q13
easy

Which models does Databricks classify as custom models for Model Serving?

Q14
medium

A team wants one serving endpoint to expose both a current production model and a challenger version while controlling request percentages. Which endp...

Q15
hard

An endpoint creator no longer has the Unity Catalog permissions needed by a served model. What must the team do to change the identity used by that en...

Q16
medium

Why is scale to zero a poor choice for latency-sensitive production endpoints?

Q17
medium

A production service calls a Databricks custom model endpoint through the REST API. What authentication practice does Databricks recommend?

Q18
easy

When sending a pandas-style JSON request to a custom model serving endpoint, which format does Databricks recommend?

Q19
easy

Which models count as custom models for Databricks Model Serving?

Q20
medium

Why is scale to zero a poor choice for a latency-sensitive serving endpoint?

Sign in to see all 24 questions

Create a free account to browse all questions — completely free during our launch phase.