DP-750: how the Azure Databricks certification maps to my SAS migration
I passed DP-750 (Implementing Data Engineering Solutions Using Azure Databricks) in September 2026. Much of it is what I do every day migrating workloads to Databricks.
by Gabriel Henrique · · 2 min read · Ler em português
I'm Gabriel Henrique, a Data Engineer at a large bank, and I lead the modernization of legacy SAS routines into Databricks for the Wealth Management segment. In September 2026 I earned the Microsoft Certified: Azure Databricks Data Engineer Associate, awarded for passing the DP-750 exam.
In this article I walk through the exam topics and show where each one shows up in my work.
What DP-750 covers
- Lakeflow Declarative Pipelines to build declarative pipelines.
- Databricks SQL for querying and analytical consumption.
- Governance with Unity Catalog.
- Performance optimization.
- CI/CD with Databricks Asset Bundles.
Lakeflow Declarative Pipelines
In a declarative pipeline, I describe the tables I want and the quality rules, and the platform handles execution order, dependencies and incremental processing. Expectations keep the quality rule written next to the table: an invalid record can raise a warning, be dropped or stop the pipeline.
This helps a lot in a SAS migration. Old routines often mix validation with transformation. Separating them makes it clear what is a business rule and what is a quality control.
Unity Catalog and governance
Unity Catalog organizes data in three levels, catalog.schema.table, and centralizes permissions and lineage. After years as the Data Owner of a data catalog, this was the topic where I felt most at home.
In practice, it answers questions that used to take real effort: who can read this table, where it comes from and who consumes it. In a bank, those answers are an audit requirement.
Canonical and derived layers
The design I use in the migration has a canonical layer, with the trusted version of each entity, and derived layers for each consumer. Today the consumers are the legacy CRM and Salesforce. Both layers live in the Lakehouse, governed by Unity Catalog.
Performance
Before the certification, I had already cut processing time on legacy ELTs by more than 80% with Spark on Databricks. The exam organizes that knowledge: Delta table layout, clustering, compacting small files and choosing the right compute for each workload.
CI/CD with Asset Bundles
With Databricks Asset Bundles, jobs, pipelines and settings become versioned files in Git, with one target per environment (development, staging and production). Promotion to production happens through a CI/CD pipeline, not by clicking in the UI.
databricks bundle validate -t dev
databricks bundle deploy -t prodThat's what I mean by productizing a pipeline end to end: from design and modeling to promotion, production reliability and monitoring.
A tip for the exam
Study by tying each topic to a real workload. Take a job you already know and ask: what would it look like as a declarative pipeline, who has permissions on it in Unity Catalog and how would it be promoted with a bundle. If you don't have the cloud basics yet, start with AZ-900.
On the portfolio of Gabriel Henrique, Data Engineer you'll find my other projects and certifications. I'm Gabriel Henrique, and I write about Databricks, governance and legacy migration.