Free Databricks ML Associate practice test

A. ✓ Correct: AutoML because it automatically builds high-quality models with minimal code using automated feature engineering and hyperparameter tuning.

B. × Incorrect: Foundation Model Fine-tuning because it involves customizing foundation models, not building them from scratch with automated tools.

C. × Incorrect: Vector Search because it stores and queries embedding vectors for RAG applications, not for model training automation.

D. × Incorrect: Distributed Training because it refers to examples of distributed deep learning using frameworks like Ray or DeepSpeed, not automating the creation of high-quality models.

A. × Incorrect: MLflow Tracking records experiments, parameters, metrics, and artifacts to manage the complete lifecycle of model development but does not prepare data for training.

B. ✓ Correct: Data preparation involves cleaning, transforming, and organizing raw data before model training.

C. × Incorrect: Model deployment involves serving models in production environments, but it does not prepare data for training.

D. × Incorrect: Feature engineering transforms raw data into useful model inputs but does not involve the initial steps of data preparation.

A. × Incorrect: This because having MLflow version 3 installed is not a necessary condition for using Unity Catalog Model Registry, although it enhances its functionality.

B. ✓ Correct: This because Unity Catalog must be enabled in your workspace to use the Unity Catalog Model Registry for managing machine learning models.

C. × Incorrect: This because using the Workspace Model Registry (legacy) is an alternative when Unity Catalog is not available, not a necessary condition.

D. × Incorrect: This because creating a new registered model is one of the actions you can perform after enabling Unity Catalog and setting up the prerequisites.

A. ✓ Correct: Unity Catalog Models because it extends the benefits of Unity Catalog to ML models, including centralized access control and auditing.

B. × Incorrect: Dedicated Compute Mode because it refers to a compute resource requirement rather than a feature for managing model lifecycle.

C. × Incorrect: MLflow Tracking Server because it is used for tracking experiments but does not provide centralized access control or auditing features.

D. × Incorrect: Workspace Model Registry because Databricks recommends Unity Catalog Models over the Workspace Model Registry for governing and deploying models.

A. × Incorrect: "MLflow Tracking is a package delivery company." does not satisfy the stem as directly as MLflow Tracking records experiments, parameters, metrics, and artifacts. and reflects a different concept.

B. × Incorrect: "Artifacts are unrelated to ML experiments." does not satisfy the stem as directly as MLflow Tracking records experiments, parameters, metrics, and artifacts. and reflects a different concept.

C. ✓ Correct: "MLflow Tracking records experiments, parameters, metrics, and artifacts." best answers the stem and aligns with the key idea behind MLflow Tracking records experiments, parameters, metrics, and artifacts.

D. × Incorrect: "Experiment metrics should never be recorded." does not satisfy the stem as directly as MLflow Tracking records experiments, parameters, metrics, and artifacts. and reflects a different concept.

A. × Incorrect: A model registry does more than just serve as a keyboard shortcut; it manages model versions and lifecycle stages.

B. × Incorrect: Lifecycle stages directly relate to the readiness of models for deployment, indicating their status in different phases like testing or production.

C. ✓ Correct: This option accurately describes how Databricks Model Registry assists with version control and stage management.

D. × Incorrect: Tracking model versions is essential for maintaining a clear history and ensuring proper deployment.

A. ✓ Correct: Using appropriate algorithms, evaluation metrics, and validation practices ensures the best possible outcome during model training.

B. × Incorrect: Ignoring validation data can lead to overfitting or underfitting of the model.

C. × Incorrect: Considering training as merely a storage account name misunderstands the purpose and process of machine learning training.

D. × Incorrect: Algorithms and metrics play a critical role in determining the effectiveness and accuracy of a trained model.

A. ✓ Correct: MLflow Tracking is used to log parameters, metrics, and artifacts during model development, ensuring reproducibility and traceability of experiments.

B. × Incorrect: Databricks Feature Store handles feature engineering tasks but does not track experiments or log parameters/metrics/artifacts.

C. × Incorrect: Unity Catalog manages data lineage and metadata for tables but does not handle tracking of ML experiments.

D. × Incorrect: AI Playground allows no-code development and testing of generative AI models, unrelated to experiment tracking.

A. ✓ Correct: The Databricks Feature Store ensures consistency between training and inference by handling all feature computation tasks, eliminating any discrepancies.

B. × Incorrect: Automatic scaling for real-time applications is a benefit of Mosaic AI Model Serving, not the Feature Store.

C. × Incorrect: Direct access to models from SQL queries is a capability of Mosaic AI Model Serving, not the Feature Store.

D. × Incorrect: Integration with Unity Catalog provides governance and lineage tracking but does not directly ensure consistent feature computations.

A. × Incorrect: Automatic scaling based on serverless compute because Mosaic AI Model Serving automatically scales up or down to meet demand changes, saving costs and optimizing latency performance.

B. ✓ Correct: Manual scaling based on user input because the service automatically handles scaling without requiring manual intervention.

C. × Incorrect: Fixed infrastructure with no scaling capabilities because Model Serving uses serverless compute which allows for automatic scaling.

D. × Incorrect: Scaling based on scheduled time intervals because Mosaic AI Model Serving scales dynamically in response to real-time demand changes.