What is the role of checkpoints in a DLT streaming inference pipeline?

They monitor the health of the pipeline and send alerts.

They store the state of the streaming query to enable fault tolerance and exactly-once processing.

They define the schema of the streaming data.

They store the trained ML model for inference.

A team has a trained scikit-learn model saved as an MLflow artifact. They need to apply it to a streaming DataFrame in DLT. The model expects a pandas DataFrame. Which approach correctly applies the model to each micro-batch?

Convert the Spark DataFrame to pandas and apply the model in the driver.

Use model.transform() directly on the Spark DataFrame.

Use a pandas UDF (applyInPandas) to apply the model to each group of rows.

Register the model as a SQL UDF and call it in a SQL query.

A data engineer is monitoring a DLT pipeline that performs streaming inference. The pipeline is processing data slower than it arrives, causing a growing backlog. The engineer wants to identify the bottleneck. Which DLT feature should they use first?

Check the cluster's auto-scaling configuration and increase the max workers.

Inspect the MLflow experiment for the model's training metrics.

Look at the Delta table's file size and optimize it with Z-ORDER.

Review the pipeline event log and the streaming query progress metrics.

A team is designing a DLT pipeline for streaming inference. The pipeline must read from Kafka, apply a model, and write predictions to a Delta table. The team also needs to ensure that the pipeline can recover from failures without reprocessing data. What is the most important configuration to set?

Enable checkpointing by specifying a checkpoint location.

Configure the cluster to use spot instances to reduce cost.

Set the trigger interval to 1 second to minimize data loss.

Use a materialized view for the output table to ensure durability.

DatabricksCertified Machine Learning Associate

Domain 4Objective 4

Identify How Streaming Inference Is Performed with Delta Live Tables MACHINE-LEARNING-ASSOCIATE Practice Questions (Page 3)

Part of the Section 4: Model Deployment domain, which makes up ~30% of our current practice bank. Databricks does not publish an official question count, but from its 90-minute exam (~35–60 total, ~11–18 in this domain), expect 2–3 from this objective — we provide 20 practice questions to prepare you well beyond it. (estimate)

20questions here

4free pages

5concepts

Questions 11–15

11foundation · easy
In a Delta Live Tables pipeline, which type of table is specifically designed to process data incrementally as new data arrives?
Select an answer first
12foundation · easy
What is the role of checkpoints in a DLT streaming inference pipeline?
Select an answer first
13application · medium
A team has a trained scikit-learn model saved as an MLflow artifact. They need to apply it to a streaming DataFrame in DLT. The model expects a pandas DataFrame. Which approach correctly applies the model to each micro-batch?
Select an answer first
14application · medium
A data engineer is monitoring a DLT pipeline that performs streaming inference. The pipeline is processing data slower than it arrives, causing a growing backlog. The engineer wants to identify the bottleneck. Which DLT feature should they use first?
Select an answer first
15expert · hard
A team is designing a DLT pipeline for streaming inference. The pipeline must read from Kafka, apply a model, and write predictions to a Delta table. The team also needs to ensure that the pipeline can recover from failures without reprocessing data. What is the most important configuration to set?
Select an answer first

Finished these 5 questions?

Review the revealed explanations, or continue through the curriculum.

Next 5 questions →Back to MACHINE-LEARNING-ASSOCIATE

Free Basic Practice is a study aid with revealable answers — not a scored exam. Examers.io is independent and not affiliated with or endorsed by Databricks. “MACHINE-LEARNING-ASSOCIATE” is a trademark of its owner, used for identification only.