
Databricks Certified Data Engineer Professional is the advanced Databricks data engineering exam. It tests production skills on the lakehouse: Lakeflow Pipelines and Auto Loader, Structured Streaming state, CDC with AUTO CDC, Unity Catalog security and governance, Liquid Clustering and Predictive Optimization, system tables, and deployment with Declarative Automation Bundles (formerly Databricks Asset Bundles). A new version of the exam was published on 9 October 2026, with new names and new topics, so older study material is out of date.
This guide follows the official exam guide dated 9 October 2026: nine sections with weights, the renamed features to know, a 7-week plan and 15 original scenario questions.
Quick facts (Databricks exam guide, 9 October 2026)
| Item | Detail |
|---|---|
| Questions | 60 scored multiple-choice questions (unscored questions may be added) |
| Time limit | 120 minutes |
| Fee | USD 200 plus applicable taxes |
| Delivery | Online proctored or test centre |
| Test aids | None allowed |
| Prerequisite | None required; related training highly recommended |
| Recommended experience | One year of hands-on data engineering on the topics in the outline |
| Validity | 2 years; recertify by passing the current exam |
Read the official Data Engineer Professional page and download the exam guide. Databricks asks candidates to recheck it two weeks before the exam date. If parts of the web page and the PDF guide differ, follow the guide dated 9 October 2026. Not ready yet? Start with the Data Engineer Associate guide.
Exam sections and weights
| Section | Weight |
|---|---|
| 1. Developing Code for Data Processing using Python and SQL | 23% |
| 2. Data Ingestion & Acquisition | 12% |
| 3. Data Manipulation | 12% |
| 4. Monitoring and Alerting | 10% |
| 5. Cost & Performance Optimization | 15% |
| 6. Ensuring Data Security and Compliance | 8% |
| 7. Data Governance | 5% |
| 8. Debugging and Deploying | 10% |
| 9. Data Modeling | 5% |
Renamed features you must recognise
| Current name (used in the exam) | Older name |
|---|---|
| Declarative Automation Bundles | Databricks Asset Bundles (DABs) |
| AUTO CDC APIs | APPLY CHANGES |
| Databricks Git folders | Repos |
| Lakeflow Pipelines / Spark Declarative Pipelines | Delta Live Tables (DLT) |
| Lakeflow Jobs | Workflows / Jobs |
The guide itself notes that AI chatbots often give outdated answers on Databricks. Check every answer against the current Databricks documentation.
1. Code for data processing (23%)
- Python project structure for Declarative Automation Bundles and CI/CD; fixing library dependency conflicts on serverless, pipelines and bundles.
- UDFs in Pandas, Python and SQL, including Unity Catalog functions.
- Streaming pipelines with Lakeflow Pipelines and Auto Loader; streaming table vs materialized view; CDC with AUTO CDC and SCD Type 1 and Type 2 (
stored_as_scd_type). - Lakeflow Jobs from the UI, API and CLI, with If/Else and For Each tasks; choosing compute, including serverless environments and performance mode.
- Tests with
assertDataFrameEqual,assertSchemaEqualandDataFrame.transform. - Structured Streaming state: watermarks, output modes,
foreachBatchand checkpoints for exactly-once recovery.
2. Ingestion (12%) and 3. Manipulation (12%)
- Ingesting Delta, Parquet, JSON, Iceberg, CSV and binary from Kafka, Kinesis, Pub/Sub and cloud storage.
- CDC from SQL Server, MySQL and PostgreSQL with Lakeflow Connect; Delta or Iceberg targets.
- OpenSharing (Databricks-to-Databricks and Databricks-to-Open), Clean Rooms and Lakehouse Federation with Unity Catalog permissions.
- Window functions, joins and aggregations in Spark SQL and PySpark; the VARIANT type with
parse_jsonandvariant_get;ai_queryfor model inference in pipelines; expectations to quarantine, drop or fail bad records.
-- Lakeflow Pipelines (SQL): SCD Type 2 customer dimension from a CDC feed
CREATE OR REFRESH STREAMING TABLE dim_customer;
CREATE FLOW customer_cdc AS AUTO CDC INTO dim_customer
FROM stream(bronze.customer_changes)
KEYS (customer_id)
APPLY AS DELETE WHEN op = 'DELETE'
SEQUENCE BY event_ts
STORED AS SCD TYPE 2;
Check the exact syntax in the current Lakeflow documentation before relying on it, since the syntax still changes between releases.
4. Monitoring (10%) and 5. Cost and performance (15%)
- System tables (billing, compute, access, lakeflow) for cost and audit; REST API, CLI and SDK; pipeline event logs; Lakehouse alerts; the Jobs UI and API.
- Unity Catalog managed tables, Predictive Optimization and Liquid Clustering (including
CLUSTER BY AUTO); deletion vectors; Delta cache. - Change Data Feed for incremental downstream processing; the query profile for data skipping, joins and shuffles; Liquid Clustering vs partitioning and ZORDER.
6. Security (8%), 7. Governance (5%), 8. Debugging and deploying (10%), 9. Modelling (5%)
- Least-privilege grants on Unity Catalog objects; ABAC policies with governed tags for row filters and column masks; hashing, tokenisation and generalisation of PII; GDPR-style deletion with Delta Lake.
- Unity Catalog tags, comments and permission inheritance.
- Spark UI, cluster logs, system tables and query profiles for diagnosis; repair runs with parameter overrides; deploying with bundles; CI/CD with Git folders.
- Delta and Iceberg table layout and compaction; dimensional models with materialized views and Unity Catalog metric views.
7-week study plan
| Week | Focus | Lab (Databricks Free Edition) |
|---|---|---|
| 1 | Spark SQL and PySpark transformations, windows, VARIANT, UDFs | Parse a JSON event feed into VARIANT and query it |
| 2 | Auto Loader, Lakeflow Pipelines, streaming tables vs materialized views, expectations | Bronze–silver–gold pipeline with quarantine rules |
| 3 | AUTO CDC, SCD Type 1 and 2, Structured Streaming state and checkpoints | SCD Type 2 dimension from a CDC feed |
| 4 | Lakeflow Jobs (If/Else, For Each), serverless compute, Lakeflow Connect, sharing and federation | Job with a For Each task over tables |
| 5 | Liquid Clustering, Predictive Optimization, CDF, query profile, system tables | Cost report from system.billing tables |
| 6 | Unity Catalog grants, ABAC, column masks, row filters, GDPR deletes | Mask a PAN column for non-finance users |
| 7 | Bundles, Git folders, CI/CD, unit tests, repair runs; review the sample questions | Deploy the pipeline with a bundle; book the exam |
15 scenario questions with answers
These are original scenarios written for this guide, not exam questions.
1. A gold aggregate must reflect late updates to past days on every refresh. Streaming table or materialized view?
Materialized view, because it maintains the full aggregate result.
2. Raw files land continuously in cloud storage and new columns appear over time. What ingests them?
Auto Loader with schema evolution, feeding a streaming table.
3. A customer dimension must keep history of address changes from a CDC feed. What do you use?
AUTO CDC with SCD Type 2.
4. Changes from an operational PostgreSQL database must replicate continuously with little code. What do you use?
A Lakeflow Connect managed connector for PostgreSQL.
5. A windowed streaming aggregation's state keeps growing because of very late events. What bounds it?
A watermark on the event-time column.
6. After a driver failure the stream must resume without double counting. What makes that possible?
The query checkpoint, which restores offsets and state.
7. Rows with a null order_id must be kept aside without stopping the pipeline. What do you use?
An expectation that quarantines (or drops) failing records instead of failing the update.
8. A job must run the same notebook for each of 30 tables. Which job feature?
A For Each task in Lakeflow Jobs.
9. Finance wants a monthly breakdown of DBU spend by job. Where do you query?
The billing system tables.
10. A large table is filtered by different columns over time and partitioning has caused small files. What do you use?
Liquid Clustering (consider CLUSTER BY AUTO) with Predictive Optimization on managed tables.
11. A downstream job needs only the rows that changed in a Delta table. Which feature?
Change Data Feed.
12. Analysts outside finance must see only the last four digits of card numbers across many tables. What scales best?
ABAC policies with governed tags that apply column masks.
13. A customer asks for their data to be erased. What must a Delta-based purge include?
Deleting the rows, then removing old file versions with VACUUM after the retention period, so the data is gone from history too.
14. A nightly job failed on one task because of a bad parameter. How do you recover quickly?
Repair the run with a parameter override, re-running only the failed and dependent tasks.
15. Pipelines, jobs and their settings must deploy the same way to dev, test and prod. What do you use?
Declarative Automation Bundles with per-target configuration, run from CI/CD.
After the Professional exam
- Compare platforms in Snowflake vs Databricks for data engineers in India.
- Add GenAI with the Databricks Generative AI Engineer Associate, or go multi-cloud with AWS DEA-C01 and DP-700.
- See live data engineering jobs.
Cloud Soft Solutions in Ameerpet, Hyderabad teaches Databricks, PySpark, Delta Lake and cloud data engineering with real-time projects. See the APEX program and all courses.
📞 Call +91 96660 19191 · WhatsApp · Flat No. 513, 5th Floor, Nilagiri Block, Aditya Enclave, Ameerpet, Hyderabad
Frequently Asked Questions
When was the current Databricks Data Engineer Professional exam published?
9 October 2026, according to the official exam guide.
How many questions are in the Databricks Data Engineer Professional exam?
60 scored multiple-choice questions in 120 minutes; unscored questions may be included.
How much does the exam cost?
USD 200 plus applicable taxes.
Which section has the most weight?
Developing Code for Data Processing using Python and SQL, at 23%.
What are Declarative Automation Bundles?
The current name for Databricks Asset Bundles (DABs), used to deploy Databricks resources through CI/CD.
How long is the certification valid?
2 years. To recertify you take the current live exam.



