The Databricks ML exams (Associate / Professional) cover the entire machine learning lifecycle: MLflow, AutoML, Feature Store, distributed training, and Model Serving. This article organizes the domain-by-domain weight tables and typical question patterns, then provides 3 domain-specific practice questions with detailed explanations.
MLA is a 48-question, 90-minute exam that tests ML implementation skills on Databricks. The current exam guide (the version as of March 1, 2025) includes registering models in the Unity Catalog registry and promoting models with aliases.
| Domain | Weight | Main Topics |
|---|---|---|
| Databricks Machine Learning | 38% | AutoML, Feature Store in Unity Catalog, MLflow (logging runs, Client API), the Unity Catalog model registry and aliases, ML runtimes |
| ML Workflows (called Data Processing in the exam guide) | 19% | Summary statistics, outlier removal, visualization, missing-value imputation, one-hot encoding, log transformation |
| Model Development | 31% | Algorithm selection, imbalanced data, estimators vs transformers, Hyperopt and SparkTrials, cross-validation, classification/regression metrics, bias-variance trade-off |
| Model Deployment | 12% | Batch vs real-time vs streaming inference, deploying to Model Serving, batch inference with pandas, splitting traffic between endpoints |
MLP is the advanced 59-question, 120-minute exam that tests your ability to design and operate ML systems at enterprise scale. The current exam guide covers the version live as of September 30, 2025, and all of its sample questions are scenarios in which you read the situation and requirements and pick the best approach.
| Domain | Weight | Main Topics |
|---|---|---|
| Model Development | 44% | SparkML pipelines, pandas Function API, distributed tuning with Optuna/Ray, data parallelism / model parallelism, nested MLflow runs, point-in-time lookups and Feature Serving in the Feature Store |
| ML Ops | 44% | Model lifecycle management, unit/integration tests, environment setup with Declarative Automation Bundles, automated retraining, drift detection with Lakehouse Monitoring (now called data profiling) |
| Model Deployment | 12% | Blue-green/canary deployment strategies, rollouts with Model Serving, serving custom PyFunc models (REST API, MLflow Deployments SDK) |
mlflow.autolog() automatically records parameters, metrics, and models for supported libraries (scikit-learn, XGBoost, LightGBM, PyTorch, etc.). A common exam pattern asks what autolog() does not record. The key gotcha is that custom business metrics (e.g., profit margin or cost-weighted scores) must still be logged manually with mlflow.log_metric().
The defining feature of Databricks AutoML is its "glass-box" approach: each trial is generated as a complete notebook. The exam frequently probes the boundary of what AutoML does automatically versus what it does not. Algorithm selection, preprocessing, and hyperparameter search are automatic; fetching external data, defining custom loss functions, and production deployment are manual.
In Feature Engineering in Unity Catalog's create_training_set(), lookup_key specifies the join key between the training data and the feature table. The MLP exam guide includes ensuring point-in-time correctness in feature lookups to prevent data leakage, which is what Point-in-Time Lookup (timestamp_lookup_key) is for: retrieving the feature values that were actually available at each row's point in time.
Hyperopt's SparkTrials parallelizes hyperparameter search across Spark workers. The MLA exam guide covers tuning with Hyperopt's fmin and parallelizing single-node models for tuning, which is where SparkTrials comes in. The MLP exam guide moves on to distributed tuning with Optuna (integrated with MLflow) and Ray, comparing Ray and Spark, parallelizing group-specific training with the pandas Function API, and choosing between data and model parallelism.
Databricks Model Serving serves custom models (Python models packaged in MLflow format) and foundation models. Foundation models are either hosted by Databricks (Foundation Model APIs) or hosted outside Databricks and accessed through External Models (for example, OpenAI). The MLA exam guide asks you to tell apart batch, real-time, and streaming inference: use a serving endpoint for low-latency real-time inference, and for large-scale batch processing, batch inference such as a Spark UDF (mlflow.pyfunc.spark_udf) or score_batch() for Feature Store models. Batch inference with ai_query() is an objective in the Generative AI Engineer Associate exam guide, not the MLA guide.
One question each from MLflow, Feature Store, and distributed training. After answering, review the explanation and the related concepts to lock in the knowledge.
MLflow
Question 1
In MLflow Model Registry in Unity Catalog, which is the recommended procedure for switching the production model to a new version?
Correct answer: B
The Unity Catalog-integrated Model Registry manages model versions with aliases instead of the legacy Staging/Production stages. The recommended flow is: set the challenger alias on the new version → validate quality via A/B or shadow testing → if successful, move the champion alias to the new version. Alias changes take effect immediately, and Serving endpoints that reference the model via an alias like models:/model_name@champion will automatically use the new version. Option A is the old Workspace Model Registry approach and is no longer recommended in the Unity Catalog-integrated version.
Feature Store
Question 2
Why would you specify timestamp_lookup_key when calling create_training_set() in Feature Engineering in Unity Catalog?
Correct answer: B
timestamp_lookup_key is the parameter that enables Point-in-Time Lookup. In time-series ML there is a fundamental data-leakage constraint: you must not use feature values from the future to predict at a given point in time. When you set timestamp_lookup_key to the event timestamp column, only the latest feature values that were available at each training row's timestamp are joined, preventing future information from leaking in. A typical example is purchase prediction, where only user features computed before the target date should be used. The MLP exam guide also includes ensuring point-in-time correctness in feature lookups.
Distributed Training
Question 3
What is the most important difference between using SparkTrials and Trials (single node) in Hyperopt?
Correct answer: B
SparkTrials is Hyperopt's parallelization backend that distributes each hyperparameter trial across Spark worker nodes. For example, running a TPE search with max_evals=100 and SparkTrials(parallelism=10) executes up to 10 trials in parallel and completes in roughly one tenth the time of a single-node run. Trials, on the other hand, runs sequentially on the driver, so all 100 trials run in series. The search algorithm (TPE, random, etc.) can be specified for either, so A is wrong. There is no GPU/CPU restriction either, so C is wrong. On Databricks, combining either with mlflow.autolog() automatically logs the results to MLflow.
MLflow is at the heart of MLA's Databricks Machine Learning domain (38%), and on MLP it appears both in Model Development (advanced MLflow usage such as nested runs) and in ML Ops (model lifecycle management). It is the single most important topic. On Free Edition (the successor to Community Edition), run through mlflow.start_run(), log_param(), log_metric(), and log_model(). Walk through the MLflow UI, observe autolog() behavior, and practice registering models and setting aliases in the Model Registry end-to-end.
Lock in the trade-offs between Accuracy, Precision, Recall, F1, and AUC-ROC for classification, and RMSE, MAE, and R² for regression. In particular, the reason Accuracy is inappropriate on imbalanced data (predicting the majority class always yields high accuracy) and the Precision-Recall trade-off (raising one by threshold tuning lowers the other) are frequent exam topics.
Be able to map each phase — data preparation → feature engineering → model training → evaluation → deployment → monitoring — to the Databricks feature you would use. Internalizing the flow Feature Store → AutoML/MLflow → Model Registry → Model Serving → Lakehouse Monitoring is what lets you make the right design call on MLP scenario questions. (Lakehouse Monitoring is now called data profiling in the Databricks documentation, but the MLP exam guide still uses the old name.)
Do the MLA and MLP exam domains overlap?
MLA (ML Associate) and MLP (ML Professional) both have domains named Model Development and Model Deployment, but the depth of the questions is very different. MLA tests whether you can perform basic machine learning tasks on Databricks: basic MLflow operations, using AutoML, Feature Store concepts, data preprocessing, and core evaluation metrics. MLP tests whether you can design and operate ML systems at enterprise scale: scalable SparkML pipelines, distributed tuning with Optuna and Ray, MLOps such as testing, automated retraining, and drift detection, and deployment strategies such as blue-green and canary. MLP assumes you already have the MLA knowledge, and the sample questions in its exam guide are practical scenarios.
What is the difference between MLflow Model Registry in Unity Catalog and Workspace Model Registry?
Workspace Model Registry manages models per workspace and uses stage transitions (Staging → Production → Archived) for lifecycle management. The Unity Catalog-integrated Model Registry lets you share models across the entire account and uses aliases (arbitrary names like champion/challenger) to reference versions instead of stages. As of 2026, Databricks recommends the Unity Catalog-integrated Model Registry, and the MLA exam guide includes both the benefits of the Unity Catalog registry over the workspace registry and promoting a challenger to champion with aliases.
How far can Free Edition take you when studying for the ML exams?
Free Edition (the successor to the retired Community Edition) lets you actually practice MLflow Tracking (experiment logging and autolog) and scikit-learn/XGBoost model training on serverless compute. However, Model Serving is limited in the number of active endpoints and has no GPU serving endpoints, and you cannot configure compute yourself, so you will need to supplement multi-node distributed training and cluster-configuration topics with official documentation and question-bank practice.
Test yourself with the ML question bank
Build pass-ready skills with 850 MLA and 1,000 MLP questions (practice sets and mock exams)
Try free questions →Related Databricks ML Articles
ML Associate Complete Guide
Full breakdown of MLA's exam scope, difficulty, and prep strategy
ML Professional Complete Guide
Full breakdown of MLP's exam scope, difficulty, and prep strategy
MLflow Complete Guide
Comprehensive coverage of Experiment Tracking and Model Registry
Free Databricks Question Bank
6,200+ bilingual practice questions
Practice with certification-focused question sets
Try free questionsNicheeLab Editorial Team
NicheeLab editorial team focused on data engineering and cloud certification learning. Content is structured around practical study needs and official exam domains.
Databricks Certifications: All 8 Exams, Difficulty & Study Plan (2026)
Complete guide to all 8 Databricks certifications — Data Eng...
Databricks Exam Difficulty Ranking: 7 Certs Compared (2026)
Seven Databricks certifications ranked by difficulty, with s...
Databricks Study Guide: Fastest Pass Route & Time Estimates (2026)
How to pass Databricks certifications efficiently. Official ...
Databricks Data Engineer Associate: Complete Guide (2026)
Domain-by-domain breakdown of the Databricks Certified Data ...
Databricks Data Engineer Professional: Complete Guide (2026)
Tactics for the Databricks Certified Data Engineer Professio...