The short version
The Databricks Machine Learning Associate exam checks that you can do basic machine learning work on the Databricks platform: explore data, build features, train and tune a model, track it with MLflow, register it in Unity Catalog, and serve it. It has 48 scored questions and a 90-minute limit. This guide follows the exam guide that Databricks lists as the version live as of March 1, 2025, and the vendor’s certification page as we read it on October 18, 2026.
What the exam is
The vendor lists it as the Databricks Machine Learning Associate certification exam. It sits at the associate level, and Databricks lists its professional-level exams separately. Databricks says that a person who passes can be expected to complete basic machine learning tasks using Databricks and its associated tools.
Two things set it apart from a general machine learning test. First, it is about the platform. You are expected to know what an ML runtime gives you, what AutoML does and does not do, how a feature table in Unity Catalog differs from an old workspace one, and what an alias on a registered model points at. Second, it is hands-on in spirit. The exam guide uses verbs like “create,” “write,” “train,” “score,” “register,” and “promote,” so expect questions that ask which call or approach fits a situation, not only what a term means.
All machine learning code on the exam is in Python. Databricks says that other code, such as data manipulation outside the machine learning steps, may be in SQL.
Who it is for
Databricks does not list a formal audience beyond the description above. In practice it suits data scientists and machine learning engineers who already train models and want to show they can do it on Databricks. It also suits data engineers moving into machine learning work. The vendor recommends about six months of hands-on experience with the tasks in the exam outline, and says course attendance is highly recommended. Nothing is required. The prerequisites page covers what to know first.
Format at a glance
| Item | What Databricks says |
|---|---|
| Level | Associate |
| Scored questions | 48, multiple choice or multiple selection |
| Unscored items | Possible, not identified, with extra time included |
| Time | 90 minutes |
| Passing score | Not published |
| Aids | None allowed |
| Delivery | Proctored; the certification page says online or at a test center |
| Languages | English, Japanese, Portuguese (Brazil), Korean |
| Validity | 2 years; recertify by taking the current exam |
| Prerequisites | None required |
The exam format page has the details on scoring, results, retakes, and rescheduling. Databricks charges a registration fee, and we do not print prices here because they change. Check the vendor page when you book.
The four sections and their weightings
| Section | Weighting |
|---|---|
| Databricks Machine Learning | 38% |
| ML Workflows (called “Data Processing” in the exam guide) | 19% |
| Model Development | 31% |
| Model Deployment | 12% |
On 48 questions, those weights work out to roughly 18, 9, 15, and 6 questions. Treat that as a rough guide, because the form may include unscored items and the vendor does not publish exact counts.
Databricks Machine Learning
The largest section, and the most platform-specific. It covers MLOps best practices, ML runtimes, what AutoML does, feature tables in Unity Catalog (creating, writing, training with, scoring with, online versus offline), and MLflow: logging, finding the best run, registering models in Unity Catalog, tags, aliases, and promoting a challenger to champion.
ML Workflows (Data Processing)
Exploring and preparing data. Summary statistics on a Spark DataFrame, removing outliers, visualizing and comparing features, imputing missing values, one-hot encoding, and knowing when a log transformation helps.
Model Development
Choosing an algorithm, handling imbalanced data, estimators versus transformers, training pipelines, tuning with Hyperopt and grid, random, or Bayesian search, cross-validation (including counting how many models a search trains), classification and regression metrics, and the bias-variance trade-off.
Model Deployment
Batch, real-time, and streaming inference, deploying a custom model to an endpoint, batch inference with pandas, streaming inference with Delta Live Tables, and splitting traffic between endpoints.
Official source
For the vendor’s own pages, see the Databricks certification page and the exam guide (PDF). They are the final authority if anything here differs. Databricks says to check the guide again two weeks before your exam, because it is updated whenever the exam changes.
More guides for DATABRICKS-ML-ASSOCIATE
- Databricks ML Associate Study Plan
- Databricks ML Associate Sample Questions
- Is the Databricks ML Associate Exam Hard?
- Databricks ML Associate Renewal and Retakes
- Is Databricks ML Associate Worth It?
- Databricks ML Associate Cheat Sheet
- Databricks ML Associate vs Professional
- GenAI vs ML Associate: Which Databricks Exam?
Next step
Read the domains explained page to see where your time should go. When you want to check yourself, our Databricks ML Associate practice tests are original practice sets organized around these four sections.
Frequently asked questions
Does the certification expire?
Yes. Databricks says it is valid for two years. To renew, you take the full exam that is live at that time.
Is the passing score published?
No. Databricks says passing scores are set through statistical analysis and can change, so it does not publish them. Be wary of any figure you see online.
Do I need a paid Databricks workspace?
The vendor does not say. Hands-on practice is strongly recommended, so use whatever workspace you can legitimately get access to.