Databricks Certified Machine Learning Professional

Know the three sections, their weightings, the 59-question/120-minute format and the registration path before you study.

• Read the official ML Professional exam guide and the summary on this page.
• Note the weightings: Model Development (44%), MLOps (44%), Model Deployment (12%).
• Confirm the $200 fee, 2-year validity, and that ~1 year of hands-on Databricks ML experience is recommended.

Build and evaluate distributed ML pipelines and choose SparkML vs single-node per use case.

• Construct SparkML pipelines (StringIndexer, OneHotEncoder, VectorAssembler) and tune/evaluate with MLlib.
• Score SparkML models for batch and streaming, and choose SparkML vs single-node for the inference type.
• Identify when SparkML is the right tool based on data volume, model and use case.

Scale training and tuning, and track complex experiments.

• Distribute training with SparkML and the Pandas Function API; parallelize group-specific training/inference.
• Tune hyperparameters at scale with Optuna (MlflowSparkStudy) and Ray; compare Ray vs Spark and parallelization strategies.
• Use nested MLflow runs, custom metrics/artifacts, and point-in-time-correct Feature Store pipelines and online tables.

Operationalize with the deploy-code lifecycle, tests and infrastructure-as-code.

• Implement the deploy-code model lifecycle across dev/test/prod and write unit + integration tests.
• Reason about which tests re-run when code/params change across the ML pipeline stages.
• Define ML assets (experiments, registered models, serving endpoints) with Databricks Asset Bundles for CI/CD.

Detect drift, monitor models and automate retraining.

• Build Lakehouse Monitoring monitors (snapshot, time-series, inference) and apply drift statistical tests.
• Define custom metrics, slice features, track performance trends, and alert on drift thresholds and endpoint health.
• Trigger automated retraining on drift/performance alerts and select top models with an appropriate metric.

Roll out models safely and serve custom models.

• Compare blue-green vs canary rollouts; implement incremental traffic on a second Served Entity with Route Optimization.
• Register custom PyFunc models with custom artifacts in Unity Catalog.
• Deploy and query custom models via the MLflow Deployments SDK, REST API or UI.