The Databricks Certified Data Engineer Professional validates that you can build, optimize, secure, and troubleshoot production data pipelines on the Databricks Data Intelligence Platform at depth. It is a 59-question, 120-minute proctored exam covering ten domains, from production Python and SQL pipeline code to cost optimization, governance, and CI/CD. This guide covers the current format, what's on it, and how to prepare.
Last updated July 2026, reflects the November 2025 exam version (10 domains).
The Databricks Certified Data Engineer Professional is an advanced certification that proves you can build, optimize, and operate production data engineering solutions on the Databricks Data Intelligence Platform: writing and debugging Python and SQL pipeline code at depth, tuning cost and performance, enforcing security and governance with Unity Catalog, and deploying through CI/CD.
It is the second rung on the Databricks data engineering ladder, one step up from the Data Engineer Associate. Passing earns a verifiable digital badge valid for two years. Where the Associate exam often rewards recognizing the right tool, the Professional exam rewards reasoning through production scenarios where several options look plausible and only one holds up under load, failure, or an audit.
There is no formal prerequisite, but Databricks assumes Associate-level knowledge as background. In practice it fits people already operating pipelines in production who want to prove depth:
Most candidates have real production experience: deploying jobs with Databricks Asset Bundles, tuning Delta tables, and working inside Unity Catalog's permission model. If you have not sat the Associate exam, that is not a blocker, but you should be comfortable with everything it covers before you start on the Professional material.
Because Databricks periodically updates the exam, always confirm the latest details in the official exam guide before you book. As of the November 2025 version:
| Questions | 59 multiple-choice questions |
|---|---|
| Time limit | 120 minutes |
| Cost | $200 USD (plus local tax) |
| Passing score | Not officially published; ~70% is the widely cited threshold |
| Delivery | Online or in person at a test center, proctored |
| Prerequisites | None required (Associate-level knowledge assumed) |
| Validity | Valid for 2 years; recertify to stay current |
| Language | English |
Figures reflect the November 2025 exam version and can change, verify against the official guide when you register.
The Professional exam guide lists ten domains, each with an explicit weighting. Our course is organised around exactly these ten units, in the order Databricks presents them:
DABs project structure, dependency management (PyPI, wheels, source archives), Python and Pandas UDFs, production pipelines with Lakeflow Declarative Pipelines and Auto Loader, job automation, CDC with the APPLY CHANGES API, control flow, and testing pipelines.
Multi-format ingestion across files, message buses, and cloud storage, plus append-only batch and streaming pipelines on Delta.
Advanced transformations (window functions, joins, aggregations) and quarantining bad data with expectations.
Delta Sharing (Databricks-to-Databricks versus the open protocol) and Lakehouse Federation for querying external systems in place.
System tables for cost, audit, and utilization; the Query Profiler and Spark UI; monitoring jobs and pipelines through event logs, the REST API, and the CLI; SQL alerts and job notifications.
Unity Catalog managed tables, deletion vectors and Liquid Clustering, data skipping and file pruning, Change Data Feed limitations for streaming, and diagnosing bottlenecks with the Query Profile.
ACLs and least privilege, row filters and column masks, anonymization and PII masking pipelines, and purging and retention with VACUUM.
Metadata, tags, and discoverability, and the Unity Catalog permission inheritance model.
Troubleshooting with the Spark UI, cluster logs, and event logs; job repair and parameter overrides; deploying with Databricks Asset Bundles and Git-based CI/CD.
Scalable data models with Delta, Liquid Clustering versus partitioning and Z-Order, and dimensional modelling for analytics.
Like the Associate exam, this one rewards reasoning over recall, so cramming facts does not get you far. The same tight Learn, Recall, Reason loop applies, now against production-depth scenarios: learn the concept, pull it back from memory, then apply it to a scenario where two options both look right.
Short, visual chapters that explain a single idea, DABs project structure, Liquid Clustering, ABAC policies, in a couple of minutes, without the wall of docs.
Active recall locks the idea in before you move on, so it is still there on exam day instead of a week later.
Scenario questions force you to choose between close options, Delta Sharing D2D versus the open protocol, Liquid Clustering versus partitioning, which is exactly how the real exam tests you.
That's how this course is built: 37 chapters across the ten exam domains, paired with practice exams that roll out unit by unit as they're finished. The Unit 1 practice exam is live now; the rest are on the way.
It is meaningfully harder than the Associate. The extra difficulty comes from production-depth reasoning: debugging code under load, weighing cost and performance trade-offs, and applying Unity Catalog's permission model correctly, not from obscure trivia. Candidates who already operate Databricks pipelines in production and study the concepts they have not touched directly tend to do well.
Most candidates need four to eight weeks of focused study. If you already hold the Associate and work daily with Databricks Asset Bundles, CI/CD, and Unity Catalog, four weeks of review is often enough. If parts of the platform (streaming, federation, ABAC) are newer to you, budget six to eight weeks and lean on hands-on practice.
Yes. The exam is 59 multiple-choice questions delivered online or at a test center with a proctor, to be completed within 120 minutes. There are no hands-on coding tasks, but many questions show code or a production scenario and ask you to pick the correct or best answer.
Yes. The Databricks Certified Data Engineer Professional credential is valid for two years. To stay certified you retake the current version of the exam before it lapses, which keeps your certification aligned with platform updates.
All 10 domains and their weighting, plus how the exam shifts from Associate.
How the two data engineering certs differ, and which one to take first.
The entry-level exam this one builds on: format, sections, and prep.
Why "DLT" is now Lakeflow Declarative Pipelines, and what it means for the exam.
Begin with DABs project structure, the foundation the rest of the code-development domain builds on. Two minutes, and it sticks.
Start Chapter 1 →