How hard is the Databricks Machine Learning Associate exam? +
The ML theory is moderate; the platform detail is what catches people. Databricks Machine Learning is the largest section at 38% and covers AutoML, Feature Engineering in Unity Catalog, MLflow tracking and the Unity Catalog model registry. Strong scikit-learn users who haven't worked with these APIs lose points, and tuning questions ask you to read Hyperopt code and count how many models a grid search with cross-validation trains.
How many questions are on the exam and how long is it? +
There are 48 scored multiple-choice questions in 90 minutes, proctored online or at a test center. The sections are Databricks Machine Learning 38%, Model Development 31%, ML Workflows 19% and Model Deployment 12%. That's just under two minutes per question, which is enough time if you know the APIs but tight if you have to reason through code you haven't seen before.
What does it cost and how long is the certification valid? +
The registration fee is $200. There are no formal prerequisites, though Databricks recommends about six months of hands-on experience with the machine learning tasks in the exam guide. The certification is valid for two years, and you recertify by passing the current version of the exam. Databricks does not list a passing percentage on the exam page.
Which outdated MLflow and Feature Store content should I avoid? +
The exam expects the Unity Catalog model registry with aliases such as champion and challenger, not the old workspace registry stage transitions - so transition_model_version_stage is now a wrong answer. Feature tables should be created with FeatureEngineeringClient in Unity Catalog, not the legacy workspace Feature Store API. Older tutorials and practice sets still teach both, so check the date on anything you study.
How long should I prepare for the ML Associate exam? +
If you already build models in Python, 3 to 6 weeks is typical, with most of that spent in a Databricks workspace using AutoML, MLflow, feature tables and the Unity Catalog registry. Work through tuning examples by hand until you can calculate how many models a given grid and cross-validation setup trains without hesitating.
Should I take the ML Associate or the ML Professional? +
Take the Associate first if you're newer to Databricks ML. It tests whether you can use the platform's core ML features correctly. The Professional goes deeper into scaling, deployment patterns, monitoring and MLOps in production. Neither requires the other, but the Associate's platform knowledge is assumed at Professional level. If you fail either, Databricks requires a 14-day wait and charges the full fee for each attempt.