The Databricks Data Engineer Associate tests building data pipelines (ingestion, transformation, Lakeflow Jobs, governance), while the Databricks Machine Learning Associate tests building models (AutoML, MLflow, training, deployment). Both cost $200, run 90 minutes, and are valid for two years. Take the one that matches your day-to-day work first.
Last updated September 2026.
| Data Engineer Associate | ML Associate | |
|---|---|---|
| Focus | Data pipelines on the lakehouse | Machine learning on the Databricks platform |
| Questions | 45 | 48 scored (unscored items may appear) |
| Time limit | 90 minutes | 90 minutes |
| Cost | $200 USD (plus local tax) | $200 USD (plus local tax) |
| Structure | 7 sections (May 2026 version) | 4 sections, 48 objectives |
| Code on the exam | PySpark and SQL | Python for all ML code; SQL only for non-ML data work |
| Passing score | Not published | Not published |
| Recommended experience | About 6 months hands-on | 6+ months hands-on |
| Validity | 2 years | 2 years |
From the official Databricks Data Engineer Associate and Machine Learning Associate exam guides. Confirm the current version before you register.
The May 2026 revision has seven sections: the Databricks platform (lakehouse, Delta Lake, Unity Catalog, compute), data ingestion and loading (Auto Loader, COPY INTO, Lakeflow Connect), transformation and modeling through the medallion layers, Lakeflow Jobs, CI/CD with Git Folders and Asset Bundles, monitoring and troubleshooting, and governance and security. It is about getting reliable data into the right tables.
Four sections cover 48 objectives. By objective count, Databricks Machine Learning is 38% (AutoML, Unity Catalog for models and features, MLflow), Model Development 31%, Data Processing 19% (EDA, missing values, encoding, splits), and Model Deployment 12%. It is about turning those tables into tracked, registered, deployable models.
Less than the shared platform suggests. Both expect you to be comfortable in notebooks, with Spark DataFrames, and with Unity Catalog's three-level namespace. The ML exam's Data Processing section touches the same data but from a feature perspective (encoding, imputation, splits), not the pipeline one.
You build or maintain pipelines, you are an analyst moving toward engineering, or you are unsure. It is the broader platform exam and the usual first Databricks credential, and it teaches the platform the ML exam assumes.
You are a data scientist or ML practitioner who already works in Databricks notebooks and rarely owns pipelines. Your strongest skills map straight onto 69% of the ML exam.
Holding both is a strong combination for ML engineers on Databricks, because it shows you can own the path from raw data to a registered model. If you plan to take both, doing the Data Engineer Associate first usually shortens the ML preparation, since the platform material in Section 1 of the ML exam will already feel familiar. Not sure where you stand? Each course has a free readiness check: Data Engineer Associate and ML Associate.
Take the one that matches your daily work: the Data Engineer Associate if you build pipelines, the ML Associate if you build models. If you are unsure or plan to take both, start with the Data Engineer Associate, because it covers the platform the ML exam assumes.
Neither is clearly harder: both are associate-level, 90-minute exams at $200. The ML Associate is narrower and assumes you already know ML concepts, while the Data Engineer Associate is broader across the platform.
No. Databricks sets no prerequisites for either exam. The Data Engineer Associate helps because it builds platform fluency, but it is not required.
Both the Data Engineer Associate and the Machine Learning Associate are valid for two years. To stay certified you pass the current version of the exam again.
Certified's Data Engineer Associate course covers all seven exam sections in 35 short chapters, with practice exams and full mocks.
Open the Data Engineer Associate course →