Databricks-ML-Pro Practice Test 5 — 25 Questions

Question 1

An ML engineer has deployed a model to Databricks Model Serving and wants to update the endpoint to serve a new model version without changing the endpoint URL or causing downtime. Which approach is correct?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Update the existing endpoint's configuration to reference the new model version and apply the update.. To update a serving endpoint to a new model version without downtime, you update the endpoint's configuration to reference the new version and apply the change. Databricks handles the rolling update, ensuring the endpoint URL remains unchanged. Other options would either cause downtime, require client changes, or rely on non-existent automatic updates.

Answer

Use MLflow's `transition_model_version_stage` to move the new version to 'Production', which automatically updates the serving endpoint.

Answer

Create a new endpoint with the new model version, then delete the old endpoint and update DNS to point to the new endpoint.

Answer

Modify the model's signature in Unity Catalog to match the new version, and the endpoint will detect the change and update automatically.

Question 2

A data scientist is using Databricks Feature Store to build a training set for a fraud detection model. They define a feature table with a primary key of `transaction_id` and a timestamp key of `event_ts`. When creating the training set with `create_training_set`, they specify `lookup_key=['transaction_id']`. The resulting training set contains features from multiple feature tables. Which statement describes how point-in-time correctness is ensured during this operation?

Accepted Answer

The timestamp key is automatically used to join features as of the timestamp of each label event, preventing data leakage from future feature values.. Point-in-time correctness in Databricks Feature Store is achieved by time-series joins using the timestamp key. When creating a training set, the system matches each label row with feature values whose timestamps are less than or equal to the label timestamp. This ensures that only historical, non-leaking features are used. The primary key alone is insufficient; the timestamp key is essential for temporal alignment.

Answer

The `create_training_set` function automatically sorts features by their creation date and selects the latest values before the label timestamp.

Answer

The primary key alone guarantees point-in-time correctness because it uniquely identifies each transaction and its associated features.

Answer

Point-in-time correctness is enforced only if the feature tables are registered with a `timestamp` column that matches the label DataFrame's index.

Question 3

A machine learning engineer is developing a model on Databricks and wants to ensure that the model's input schema is enforced during inference. They are using MLflow to log the model. What should they do?

Accepted Answer

Log the model with `mlflow.pyfunc.log_model` and provide a `signature` that includes the input schema.. MLflow's model signature defines the expected input and output schema. When a model is logged with a signature, MLflow can validate input data during inference, ensuring that the schema is enforced. This is the standard and most effective way to enforce input schema, as it leverages built-in functionality without custom code.

Answer

Log the model with `mlflow.pyfunc.log_model` and include a custom `predict` method that validates the input schema.

Answer

Log the model with `mlflow.sklearn.log_model` and set the `input_example` parameter to a sample of the training data.

Answer

Log the model with `mlflow.sklearn.log_model` and set the `registered_model_name` parameter to register the model in the Model Registry.

Question 4

A team is deploying a model to Databricks Model Serving and wants to implement a canary release strategy to gradually shift traffic from the current model version to a new version. Which TWO configurations are required to achieve this? (Choose two.)

Accepted Answer

Create a serving endpoint with multiple served entities, each referencing a different model version.. Canary releases in Databricks Model Serving are achieved by configuring a single endpoint with multiple served entities, each pointing to a different model version. Traffic is then split between these entities using the endpoint's traffic configuration, which can be updated via the REST API. This allows gradual shifts and monitoring.

Answer

Deploy two separate endpoints and use an external load balancer to distribute traffic between them.

Answer

Enable automatic canary deployment by setting a flag in the model's MLflow metadata.

Answer

Configure the endpoint to use a single served entity and rely on the model's internal logic to route requests based on a random seed.

Question 5

A team wants to deploy a scikit-learn model to a real-time REST endpoint on Databricks. They have logged the model with MLflow and registered it in Unity Catalog. Which method should they use to create the serving endpoint?

Accepted Answer

Use the Databricks Model Serving UI or REST API to create an endpoint and select the registered model version.. Databricks Model Serving is the managed solution for deploying registered models as REST endpoints. It handles infrastructure, scaling, and dependency management automatically. Using the UI or REST API to create an endpoint from a Unity Catalog model version is the correct and supported approach, unlike manual containerization or local serving.

Answer

Package the model into a Docker image and deploy it to a Kubernetes cluster managed by Databricks.

Answer

Write a Databricks job that loads the model and exposes it via a Flask app on a driver node.

Answer

Use MLflow's `mlflow models serve` command on a Databricks cluster to start a local REST server.

Question 6

A machine learning engineer has deployed a model to a Databricks Model Serving endpoint. The model requires a custom Python package that is not available in the default environment. The engineer has already logged the model with MLflow and included the package in the conda environment. However, upon deployment, the endpoint fails to start. What is the most likely cause?

Accepted Answer

The custom package is not available in a public PyPI repository, and no additional index URL was specified.. When deploying a model with custom dependencies, Databricks Model Serving uses the conda environment logged with the MLflow model. If the package is not on PyPI or requires a private index, you must specify the index URL in the conda environment. Without it, the installation fails, and the endpoint cannot start. Ensuring the environment specification includes all necessary sources is critical for successful deployment.

Answer

The model's MLflow flavor is not supported by Model Serving.

Answer

The custom package is not installed on the cluster used for serving.

Answer

The endpoint's workload size is too small to install the package.

Question 7

Refer to the exhibit. A developer wants to ensure the Random Forest model can be used for automated inference at scale. Based on the provided code, what is missing to enable the model to support the 'predict' method within the Databricks Model Serving environment?

Accepted Answer

The model.fit(X_train, y_train) method must be called before log_model.. The provided code logs an unfitted model instance. MLflow cannot serialize a model's 'predict' capabilities until the model has actually been trained (fitted) on a dataset. Attempting to deploy or run an unfitted model results in an error, as there is no learned logic to execute. The model must be fitted with data prior to being passed to log_model.

Answer

The model must be registered to the model registry before logging.

Answer

The model must be wrapped in an mlflow.pyfunc.PythonModel class.

Answer

The code must explicitly define the input signature in log_model.

Question 8

A data science team at a retail company has registered a demand forecasting model in the MLflow Model Registry. The model version is currently in the 'Staging' stage and has been validated by the QA team. Before promoting it to 'Production', the ML engineer wants to ensure that the model's performance does not degrade when serving live traffic. They decide to deploy the model to a small percentage of production traffic while continuing to serve the existing model. Which Databricks feature should they use to achieve this?

Accepted Answer

Databricks Model Serving with traffic splitting. Databricks Model Serving natively supports traffic splitting, which lets you distribute incoming requests across multiple model versions on the same endpoint. This is ideal for canary deployments: you can send a small fraction of traffic to the new version, monitor metrics, and then increase the percentage. Other options lack the real-time routing capability required to test a model under live production load without affecting all users.

Answer

Databricks Jobs with conditional task execution

Answer

MLflow Model Registry webhooks

Answer

MLflow Projects with Docker environments

Question 9

A fraud-detection model is registered in Unity Catalog and deployed to a Model Serving endpoint. Compliance requires that every prediction be traceable to the exact model version and the request that produced it. Which combination of Databricks features should you configure to meet this requirement?

Accepted Answer

Enable inference table logging on the endpoint and register the model version in Unity Catalog with a descriptive comment.. Traceability of individual predictions requires server-side capture of the request, response, and serving model version, which is what inference tables provide, combined with Unity Catalog's governed model version identity. Training-time logging, endpoint configuration snapshots, and client-side logs each capture only part of the picture and cannot reconstruct the full per-prediction audit trail.

Answer

Attach a model signature to the registered model and enable request logging in the serving client application.

Answer

Configure the endpoint with a scale-to-zero policy and log the endpoint configuration to a Delta table daily.

Answer

Enable MLflow autologging in the training notebook and set the model version stage to Production.

Question 10

A financial institution is using Databricks to build and deploy a credit risk model. The model must comply with regulations that require full auditability of the model's lineage, including data sources, transformations, and training parameters. The team uses MLflow for tracking and the Feature Store for feature management. Which TWO of the following practices are essential to meet the auditability requirements? (Choose two.)

Accepted Answer

Use Databricks Feature Store to create feature tables and log the training set with the model, including feature lookup keys.. To meet auditability requirements, it is essential to automatically capture data and feature lineage. Logging data versions and feature table versions in MLflow provides a record of the exact inputs used. Using the Feature Store to log the training set with the model ensures that feature lineage is preserved. Together, these practices create a transparent and verifiable chain from data to model, which is necessary for regulatory compliance.

Answer

Enable MLflow autologging for all supported frameworks to capture parameters and metrics automatically.

Answer

Manually document the data sources and transformations in a separate spreadsheet for each model version.

Answer

Store all training code in a Git repository and commit after each run.

Question 11

You are implementing a CI/CD pipeline for a machine learning model on Databricks. The pipeline must automatically retrain the model when new data arrives, validate its performance, and promote it to production if it meets quality thresholds. Which TWO of the following steps are essential to include in the pipeline to ensure safe and automated deployment? (Choose two.)

Accepted Answer

Implement automated tests that evaluate the model on a holdout dataset and compare metrics against a baseline before promotion.. The essential steps for an automated CI/CD pipeline include registering the model in MLflow Model Registry with stage transitions and implementing automated tests on a holdout dataset to validate performance against a baseline. These steps ensure that only models meeting quality thresholds are promoted, enabling safe automation. Other steps like manual review or storing artifacts in Git are not suitable for automated pipelines.

Answer

Manually review the model's performance metrics in a notebook before allowing the pipeline to proceed.

Answer

Store the model artifacts in a Git repository alongside the code to ensure versioning.

Answer

Configure the pipeline to directly overwrite the production model endpoint with the newly trained model without validation to reduce latency.

Question 12

You are responsible for a fraud detection model deployed to a Databricks Model Serving endpoint. The model was trained on transaction data from the past 6 months. After two months in production, you notice a gradual decline in precision and recall. You suspect data drift. Which approach should you use to monitor feature drift for this endpoint?

Accepted Answer

Enable inference logging on the endpoint and use Databricks Lakehouse Monitoring to compare production feature distributions against the training baseline.. Inference logging captures the actual feature values sent to the model endpoint, and Lakehouse Monitoring provides built-in drift metrics by comparing these to a baseline profile. This combination enables automated, scalable drift detection without manual intervention. Other options either misrepresent platform features or fail to statistically compare distributions.

Answer

Set up a Databricks SQL dashboard that queries the endpoint's logs and calculates the average prediction value over time.

Answer

Configure the endpoint to automatically retrain the model when drift is detected using the built-in auto-retrain feature.

Answer

Use MLflow Tracking to log the model's predictions and manually compare them to ground truth labels in a notebook.

Question 13

Your team is using MLflow Model Registry to manage a model that predicts customer churn. A new version has been registered and passed validation. You need to transition this version to the 'Production' stage and ensure that all downstream scoring jobs automatically use it. What should you do?

Accepted Answer

Transition the model version to 'Production' and update the downstream jobs to reference the model by stage 'Production'.. Transitioning the model version to the 'Production' stage and having downstream jobs reference the model by stage 'Production' is the correct approach. This allows you to promote new versions to production without modifying job code, ensuring that all consumers automatically use the latest approved model. It also maintains a clear audit trail and governance.

Answer

Keep the model version in 'Staging' and update the downstream jobs to reference the model by stage 'Staging'.

Answer

Archive the current production model and register the new version with a new model name.

Answer

Transition the model version to 'Production' and update the downstream jobs to reference the model by version number.

Question 14

A data scientist has registered a model in Unity Catalog under the name `prod.ml.forecast_model`. They now need to define a service-level objective (SLO) that automatically monitors the model's prediction quality in production, not just endpoint uptime. Which Databricks feature should they configure to detect model performance degradation?

Accepted Answer

Lakehouse Monitoring on the inference table. Lakehouse Monitoring on the inference table is the only option that provides statistical monitoring of model inputs and outputs. It computes drift and profile metrics and, when ground truth labels are available, model quality metrics such as accuracy or RMSE. This makes it the correct mechanism for an SLO tied to prediction performance rather than simple endpoint availability.

Answer

MLflow Model Registry webhooks

Answer

Cluster autoscaling policies on the serving cluster

Answer

Model Serving endpoint health checks

Question 15

You are using Databricks Feature Store to manage features for a model. A feature table is updated daily with new data. Your model training job reads from the feature table and logs the model with MLflow. To ensure that the model in production uses the correct feature values at inference time, what must you do when logging the model?

Accepted Answer

Log the model with the feature store lookup tables and specify the feature table names.. To ensure the model uses the correct feature values at inference time, you must log the model with the feature store lookup tables and specify the feature table names. This enables automatic feature retrieval, maintaining consistency between training and serving and ensuring that daily updates are reflected in predictions.

Answer

Log the model with a static copy of the feature values used during training.

Answer

Log the model with the feature table's primary key and timestamp columns only.

Answer

Log the model without any feature store metadata and manually join the feature table in the scoring job.

Question 16

A team notices that a deployed Model Serving endpoint occasionally returns stale predictions for a subset of customers shortly after the nightly feature refresh completes. The offline feature table is confirmed current. Which is the most likely explanation?

Accepted Answer

The online feature store was not republished after the nightly refresh, so lookups return the previous feature values for those customers.. Feature Store keeps an offline table for training and an online table for serving. Updates to the offline table do not propagate automatically; a publish step materializes them into the online store. When the nightly refresh runs without a subsequent publish, the endpoint keeps reading the prior values, producing stale predictions for affected customers until the online store is refreshed.

Answer

The offline feature table's Delta transaction log has not been checkpointed, so readers see an older snapshot.

Answer

The model version serving traffic is older than the latest registered version, so predictions lag behind the current model.

Answer

The endpoint's autoscaling added replicas that cached outdated feature values during scale-out.

Question 17

A data scientist has trained a scikit-learn model and logged it with MLflow. They now want to register this model in the MLflow Model Registry and transition it to 'Production' to be served via Databricks Model Serving. Which of the following is a prerequisite for registering the model?

Accepted Answer

The model must be logged using the MLflow Tracking API and have a valid run ID.. Registering a model in MLflow Model Registry requires that the model artifact is linked to an MLflow run. The run ID establishes the model's origin and enables versioning. Other options are either optional or incorrect; signatures, experiment associations, and admin approvals are not required for registration.

Answer

The model must be logged to an MLflow experiment that is associated with a registered model.

Answer

The model must be logged with a signature.

Answer

The model must be approved by a workspace admin.

Question 18

A data scientist at your company has trained a scikit-learn model and logged it with MLflow using the default 'sklearn' flavor. The model must now be deployed to a Databricks Model Serving endpoint for real-time inference, and the endpoint must automatically reload the new version whenever the model's stage changes to 'Production' in the MLflow Model Registry. Which deployment approach should the team use?

Accepted Answer

Register the model version in the MLflow Model Registry, then create the serving endpoint by specifying the model name and the 'Production' stage in the endpoint configuration.. The requirement is automatic reload when a model version enters the 'Production' stage. Databricks Model Serving supports this directly by referencing the registered model name and stage in the endpoint configuration. Other approaches either bypass the registry, require manual redeployment, or rely on mechanisms the platform does not provide, making them unsuitable for a managed real-time endpoint with stage-driven updates.

Answer

Log the model to a new MLflow experiment and configure the endpoint to poll the experiment's artifact location for changes.

Answer

Package the model into a Docker image with a Flask app and deploy it to a Databricks cluster running as a job.

Answer

Export the model artifact to DBFS as a pickle file and create a Python script that loads the pickle at endpoint startup.

Question 19

A team runs a weekly retraining job that produces a new model version in Unity Catalog. Their production endpoint is currently serving version 4. They want to promote version 5 with zero downtime and the ability to roll back instantly if error rates rise. Which approach best meets these requirements?

Accepted Answer

Configure the serving endpoint to serve version 5 and rely on the platform's rolling update so traffic shifts gradually while the previous configuration remains available for rollback.. Model Serving supports rolling updates that keep the endpoint live while a new model version loads, and the previous configuration remains restorable. Repointing the endpoint to the newer version and letting the rolling update run achieves zero-downtime promotion with a straightforward rollback. Deleting the old version or duplicating endpoints introduces risk or cost without improving the outcome.

Answer

Delete version 4 from the registry, deploy version 5, and recreate the endpoint so it picks up the new version automatically.

Answer

Set the endpoint's served entities to version 5 with a traffic percentage of zero, then raise traffic to one hundred percent after validation.

Answer

Create a second endpoint for version 5, run both endpoints in parallel indefinitely, and split traffic manually at the load balancer.

Question 20

A team is preparing to promote a new model version to production in the MLflow Model Registry. They must ensure the model can be served with a consistent environment across staging and production and that dependency drift is detected before promotion. Which TWO practices should they follow? (Choose two.)

Accepted Answer

Log the model with its conda environment and requirements files so the exact dependency versions are captured as part of the model version.. Capturing the exact dependency environment with the model version and then validating that environment against the target serving environment before promotion together ensure consistent serving and catch drift early. Floating to latest versions, mismatching runtimes, or relying on bare pickles all break reproducibility or hide drift.

Answer

Store only the model's pickle file and reconstruct dependencies manually on each environment, since pickle serialization is environment-independent.

Answer

Rely on the serving endpoint to install the latest available versions of each library at deployment time so the environment stays current.

Answer

Pin the serving cluster's runtime to a different Databricks Runtime version than training used, to validate cross-version compatibility during staging.