Databricks ML Professional exam format
Here is what Databricks publishes about the Machine Learning Professional exam.
| Certification | Databricks Certified Machine Learning Professional |
| Scored questions | 59 |
| Duration | 120 minutes |
| Question format | Multiple choice |
| Registration fee | $200 |
| Validity | 2 years, then recertification on the current version |
| Prerequisites | None (training highly recommended) |
| Recommended experience | 1+ years hands-on with the exam guide’s ML tasks |
| Delivery | Online proctored or test centre |
| Languages | English only |
| Test aids | None permitted |
| Passing score | Not published by Databricks |
Databricks notes that unscored items may be included for statistical purposes and are not identified to candidates.
The 44/44/12 split
| Section | Weight | Approx. scored questions |
|---|---|---|
| Model Development | 44% | ~26 |
| ML Ops | 44% | ~26 |
| Model Deployment | 12% | ~7 |
Counts derived from Databricks’ percentages against 59 scored questions.
This is the most lopsided structure of any exam covered here, and it is lopsided in a way people do not expect.
Deployment is only 12%. Candidates see “Professional” and “production” and assume serving infrastructure dominates. It does not — about seven questions.
Two 44% sections means no safe ground. Each is worth 26 questions. Being strong in one and weak in the other leaves you needing near-perfection in the strong half, which is not a realistic plan.
The practical reading: divide your study time roughly in half between model development and ML Ops, then give deployment a few focused hours.
Two minutes per question
59 questions in 120 minutes. The same effective pace as the Data Engineer Professional, and comfortable for multiple choice.
English only
Alone with the Data Analyst Associate among Databricks certifications, this exam is offered only in English. The ML Associate, by contrast, is available in four languages. Worth planning for if English is not your first language.
The passing score is not published
Databricks does not publish one for this exam. Aim for consistent 80%+ on realistic practice material.
What “professional” means here
Not seniority in the abstract — a specific shift in emphasis. The ML Associate asks whether you can use the platform to build a model. The Professional asks whether you can:
- develop models rigorously, with proper validation, tracking and reproducibility
- operate them: monitor for drift, retrain, manage versions, handle the lifecycle
- and deploy them appropriately, though that is the smallest part
The one-year experience expectation reflects that. Much of the ML Ops section is about problems that only appear after a model has been live for months.
What the weighting means for revision
Split your time. Not by section count — by weight.
Half your study belongs in model development: experiment tracking, validation strategy, tuning, evaluation, reproducibility. The other half belongs in ML Ops: monitoring, drift, retraining, versioning, lifecycle, automation.
Deployment gets a few hours. Know the inference patterns, know how a registered model becomes a served one, and stop.
A useful check while studying: if you cannot describe how you would detect that a live model has quietly degraded, you are not ready for 44% of this exam.
Booking
Book through Databricks. Nothing gates registration, and the Associate is not a prerequisite.