ML-PROFESSIONAL practice questions 1–10

1. Question 1 · 88 in bank

   A Machine Learning Engineer has created a custom PyFunc model wrapper for a fraud detection system that needs to be registered in Unity Catalog under the schema risk_models in the production catalog. The model requires specific dependencies and must be accessible for governance and version control. The engineer is working on a dedicated cluster with Unity Catalog enabled, but the model registration is failing. Why is the model failing to be registered in this case?

   • AThe model signature must be provided, as Unity Catalog requires all models to have signatures.
   • BThe cluster must be configured with shared access mode instead of dedicated mode.
   • CThe PyFunc model must inherit from mlflow.sklearn.Model instead of mlflow.pyfunc.PythonModel.
   • DThe model artifacts must be stored in DBFS rather than Unity Catalog volumes.

   Show answer

   Answer: B

   Next question

2. Question 2 · 88 in bank

   A machine learning engineer wants to deploy a model for real-time serving using MLflow Model Serving. For the model, the machine learning engineer currently has one model version in each of the stages in the MLflow Model Registry. The engineer wants to know which model versions can be queried once Model Serving is enabled for the model. Which of the following lists all of the MLflow Model Registry stages whose model versions are automatically deployed with Model Serving?

   • AStaging, Production, Archived
   • BProduction
   • CNone, Staging, Production, Archived
   • DStaging, Production
   • ENone, Staging, Production

   Show answer

   Answer: D

   Next question

3. Question 3 · 88 in bank

   Which of the following is a reason for using Jensen-Shannon (JS) distance over a Kolmogorov-Smirnov (KS) test for numeric feature drift detection?

   • AAll of these reasons
   • BJS is not normalized or smoothed
   • CNone of these reasons
   • DJS is more robust when working with large datasets
   • EJS does not require any manual threshold or cutoff determinations

   Show answer

   Answer: E

   Next question

4. Question 4 · 88 in bank

   Which of the following deployment paradigms can centrally compute predictions for a single record with exceedingly fast results?

   • AStreaming
   • BBatch
   • CEdge/on-device
   • DNone of these strategies will accomplish the task.
   • EReal-time

   Show answer

   Answer: E

   Next question

5. Question 5 · 88 in bank

   A machine learning engineer has developed a model and registered it using the FeatureStoreClient fs. The model has model URI model_uri. The engineer now needs to perform batch inference on the training set logged with the model, but a few of the feature values in the column spend have since been updated and arc present in the customer-level Spark DataFrame spark_df. The customer_id column is the primary key of spark_df and the training set used when training and logging the model. Which code block can be used to compute predictions for the training set while overwriting its old spend values with the new spend values from spark_df?

   • Afs.score_batch(model_uri, spark_df)
   • Bfs.score_model(model_uri, spark_df)
   • Cdf = fs.get_updated_feature(spark_df, model=uri) fs.score_batch(model_uri, df)
   • Ddf = fs.get_updated_features(spark_df) fs.score_batch(model_uri, df)

   Show answer

   Answer: A

   Next question

6. Question 6 · 88 in bank

   A Machine Learning Engineer wants to monitor the quality and stability of their machine learning model’s predictions over time. They have a Delta table, retail_inference_log, which records each model prediction along with input features, a timestamp, and (when available) the true label. They need to detect data drift and monitor model performance trends using Databricks Lakehouse Monitoring, ensuring that alerts are triggered if the distribution of predictions or input features changes significantly. Which approach will set up monitoring for this use case?

   • ACreate a monitor with the Inference profile on the retail_inference_log table, specifying the timestamp column and the columns for model inputs, predictions, and labels. Configure the monitor to compute drift and performance metrics over time windows.
   • BCreate a monitor with the Inference profile on the retail_inference_log table, and specify a recent batch of production data as the baseline table for drift detection. Use this recent production data to compare against new data for drift and performance monitoring.
   • CCreate a monitor with the Snapshot profile on the retail_inference_log table, so that metrics are calculated over the entire table each time the monitor runs and therefore is able to compare new values with previous ones to compute data drift.
   • DCreate a monitor with the Time Series profile on the retail_inference_log table, specifying the timestamp column and including model input, prediction columns and the true label column. This will track drift in features and predictions over time, and model performance could also be tracked using a custom metric.

   Show answer

   Answer: A

   Next question

7. Question 7 · 88 in bank

   Which is a benefit of logging an input example with an MLflow model?

   • AThe model evaluation metrics are stored with the model
   • BThe environment requirements are packaged with the MLmodel file
   • CAn example record is available to test serving applications
   • DAn example prediction is available to test post-inference pipelines

   Show answer

   Answer: C

   Next question

8. Question 8 · 88 in bank

   Which of the following is a simple statistic to monitor for categorical feature drift?

   • AMode
   • BNone of these
   • CMode, number of unique values, and percentage of missing values
   • DPercentage of missing values
   • ENumber of unique values

   Show answer

   Answer: C

   Next question

9. Question 9 · 88 in bank

   A machine learning engineer wants to load the data from the very first version of a Delta table from location path. Which of the following lines of code can be used to accomplish this task?

   • Aspark.read.format("delta").option("version", 0).load(path)
   • Bspark.read.format("delta").option("versionAsOf", 0).load(path)
   • Cspark.read.format("delta").option("versionAsOf", 1).load(path)
   • Dspark.read.format("delta").option("version", 1).load(path)

   Show answer

   Answer: C

   Next question

10. Question 10 · 88 in bank

    A Machine Learning Engineer needs a continuous deployment pipeline for their models hosted on Databricks Model Serving. The deployment automation should execute after a model is trained and registered using MLflow. The goal of the automation is to deploy the latest version of the model from the MLflow Model Registry only if the model can meet the company’s strict latency requirements (P95 < 300ms) while serving production traffic. How can the engineer validate that new models meet their latency requirements when served in production?

    • AA/B test the latest model with Databricks Model Serving so that the latest model receives 5% of production traffic and the current model receives the rest. Use inference tables to calculate P95 latency and verify it is less than 300ms.
    • BServe the latest model on Databricks Model Serving and use a load testing client to generate requests at the production request rate. Use the load testing client to calculate P95 latency and verify it is less than 300ms.
    • CUse the MLflow Get Run API to retrieve the model metrics from MLflow Tracking and verify that the model_latency metric is less than 300ms.
    • DUse the MLflow Get Run API to retrieve the model metrics from MLflow Tracking and verify that the inference_latency metric is less than 300ms.

    Show answer

    Answer: A

    Next question

Keyboard:J Kmove between questionsA–Dselect an answerRshow or hide the answer← →change page

ML-PROFESSIONAL practice, question 1. 88 questions available.