Databricks Code Practice

Get fluent in Databricks by typing, not watching.

148 exercises, 4 production-grade pipeline labs, and 2 deep-dives. All on Databricks Free Edition.

Clone once, import into Databricks, pick a folder. Exercises fail loud until your code is right; labs ship with synthetic data so you build production-style pipelines, not toy ones.

---

Author

Jakub Lasak - Databricks Data Engineer (ex-Uber). I write about the Databricks work that gets you hired and promoted.

• LinkedIn (16.5K followers) - Databricks projects and tips
• Substack - Newsletter for data engineers, 5.5K subscribers
• Podcast - Databricks and data engineering, on Spotify, Apple, and YouTube
• DataEngineer.wiki - Interview prep and free Databricks practice

Prepping for interviews? Writing code is one half of the battle - knowing the questions that actually come up is the other. Databricks Interview Prep is where I keep the questions and answers for junior, mid and senior Databricks interviews, all levels in one place.

What's Inside

Fluency comes from reps, not reading. Three structured paths:

• exercises/ - focused reps on a single concept. LeetCode-style, 5-20 min each.
• pipeline-labs/ - end-to-end medallion pipelines on a business scenario. 2-3 hours each.
• deep-dives/ - go deep on a single topic, hands-on, end to end. 1-3 hours each.

	Exercises	Pipeline Labs	Deep-Dives
Format	Single notebook, one TODO per exercise	Multi-notebook guided project	Single-topic deep investigation
Time	5-20 min per exercise	3-8 hours per lab	1-5 hours
Scope	One concept (MERGE, window functions, ...)	End-to-end project (ingestion -> bronze -> silver -> gold)	One topic, hands-on in depth
Narrative	None. "Given table X, write..."	Business scenario. "You're building a streaming pipeline for..."	Focused. "Go deep on one topic, end to end."
Order	Pick any, skip around	Sequential notebooks that build on each other	Sequential; each step layers on the last
Goal	Drill a skill until it's automatic	See how concepts fit in a real project	Build real, hands-on command of one topic

Catalog

Exercises (exercises/)

Topic	Notebooks	Exercises	Description
Delta Lake	6	51	MERGE operations, time travel, schema enforcement, OPTIMIZE, liquid clustering, change data feed
ELT	7	53	Spark SQL joins, window functions, PySpark transformations, Auto Loader, batch ingestion, medallion architecture, complex data types
Streaming	6	44	Structured Streaming basics, windowed aggregations & watermarks, stream-static joins, stream-stream joins, foreachBatch patterns, checkpointing & recovery

Total: 19 notebooks, 148 exercises

More exercise topics coming - next up: Unity Catalog, Performance, and Lakeflow Spark Declarative Pipelines (formerly DLT).

Pipeline Labs (pipeline-labs/)

Multi-notebook, end-to-end medallion pipelines with a business scenario. Each runs 2-3 hours and ships with a synthetic data generator.

Lab	What You Build	Focus
Apparel Retail 360 (Lakeflow)	End-to-end retail analytics pipeline on Lakeflow Spark Declarative Pipelines with a full medallion architecture.	Lakeflow Spark Declarative Pipelines, Medallion, SCD Type 2, Streaming, Data Quality Expectations
Fintech Transaction Monitoring	Real-time fraud-monitoring pipeline for a payment processor handling 500K+ transactions/day.	Structured Streaming, Rescued Data, Watermarked Dedup, Stream-Static Joins, Liquid Clustering
DE Associate Certification Prep	Production-grade pipeline drilling the core topics of the Databricks Data Engineer Associate cert.	Auto Loader, COPY INTO, Medallion, SCD2, Jobs, Unity Catalog
PySpark Developer Cert Prep	E-commerce analytics pipeline covering six of the seven Spark Developer Associate exam sections.	DataFrame API, Structured Streaming, Data Skew, Performance Tuning

Deep-Dives (deep-dives/)

Single-topic labs that go deep on one technique or capability - hands-on, end to end.

Lab	What You Build	Focus
6 Delta Optimization Techniques	Measure files and bytes read on partitioned, Z-Ordered and Liquid Clustered copies of a 10M-row, 2,000-file sales table, then run OPTIMIZE, auto compaction and VACUUM and read what each one changed in the table history.	Partitioning, Z-Order, Liquid Clustering, OPTIMIZE, Auto Compaction, Deletion Vectors, VACUUM
Build a Genie Agent	Stand up a Databricks Genie Agent on a retail star schema (joins, SQL expressions, 8 example queries, 3 Unity Catalog functions), score it with a 10-question benchmark, then embed it in an AI/BI dashboard.	Genie Agent, Unity Catalog Functions, Star Schema, Benchmarks, AI/BI Dashboards

How to Use

1. Sign up for Databricks Free Edition (free, no credit card)
2. Clone or import this repo into Databricks (Workspace -> Create -> Git folder)
3. Open the folder you want. For exercises, open any numbered notebook in a topic folder (it sets up its own data). For pipeline labs and deep-dives, open the lab's README and follow it.

Everything runs on Free Edition: serverless compute, Unity Catalog, Delta Lake. No cloud account, no cluster config.

Which Should I Start With?

• New to Databricks? Start with DE Associate Cert Prep - broadest fundamentals.
• Want quick reps on a specific concept? Delta Lake exercises, ELT exercises, or Streaming exercises - drill one concept at a time.
• Comfortable with batch, new to streaming? Start with Streaming exercises for atomic concept drills, then Apparel Lakeflow and Fintech Monitoring for end-to-end pipelines.
• Preparing for a cert? DE Associate or Spark Developer Associate.
• Already shipping pipelines, want to go deeper on performance? Delta Optimization Techniques.
• Want to ship natural-language analytics (Genie / AI/BI)? Build a Genie Agent - connect Genie to a star schema, benchmark its answers against ground-truth SQL, embed it in a dashboard.

Stay in the Loop

New exercises and labs ship regularly. Follow on LinkedIn or subscribe to the Substack newsletter to be notified when new content drops.

Changelog

• 3 June 2026 - New deep-dive: Build a Genie Space - natural-language analytics on a retail star schema, with a regression benchmark and an embedded AI/BI dashboard.
• 2 June 2026 - New exercise topic: Streaming - 45 LeetCode-style reps across 6 notebooks (Structured Streaming, watermarks, stream-static & stream-stream joins, foreachBatch, checkpointing).
• 18 April 2026 - Consolidated into one repo: 4 pipeline labs and the first deep-dive (Delta optimization) joined the original Delta Lake + ELT exercises.
• 7 April 2026 - Initial release: Delta Lake + ELT exercises.

Feedback

Found an exercise that can't be solved as written? That's the most useful thing you can report - open an issue. Suggestions for new topics welcome too.

Finished an exercise and want to keep your work? Fork the repo - your solutions stay in your fork. Please don't open a PR with solved TODO cells; it deletes the exercise for everyone else. See CONTRIBUTING.md.

License

CC BY-SA 4.0. In plain terms:

• Use it, fork it, adapt it, teach from it - at work, in a study group, on your blog, in a bootcamp, in paid training. Commercial use of the material is fine (Free Edition itself is for non-commercial use, so paid training runs on a paid Databricks workspace).
• Credit it clearly. Name this repo and link back, visibly enough that your audience knows where the exercises came from. Don't pass them off as your own material.
• Share adaptations under the same terms. If you publish a modified version, it carries this license too - so nobody can republish the exercises under stricter terms.

Using this in paid training or a course? Go ahead - just keep the attribution obvious. I'd genuinely like to hear about it: say hi.

---

Disclaimer: This is an independent educational resource created by Jakub Lasak. Not affiliated with, endorsed by, or sponsored by Databricks, Inc. "Databricks" and "Delta Lake" are trademarks of their respective owners.