GitLab · Primly Community

GitLab data engineer interview: pipelines and SQL were the whole show

de_derek · 6 replies

Just finished my GitLab DE loop last month. Sharing this because I couldn't find a single recent post when I was prepping, so here's the full picture.

The process was: recruiter screen, technical phone screen, then a 4-round virtual onsite. No take-home, which I actually appreciated.

Phone screen About 45 minutes with a senior DE on the team. Half was background, half was live SQL. Not tricky stuff, but not trivial either: window functions, CTEs, a question about deduplication logic. They asked how I'd handle late-arriving events in a streaming context. Know your basics cold.

Onsite (4 rounds, same day)

Round 1: SQL deep dive. Median salary query, ranking with ties, a join-vs-subquery efficiency question. They wanted me to talk through my reasoning, not just produce the answer.

Round 2: Data modeling. Design a schema for tracking CI/CD pipeline runs. Makes total sense given what GitLab actually does. They wanted to see how I handle slowly changing dimensions and what I'd denormalize for query performance.

Round 3: Systems design. Design a near-real-time data pipeline that ingests GitLab events (commits, MR opens, deployments) and feeds a dashboard. Kafka came up, so did dbt and Snowflake. They seemed fine with BigQuery too. The interviewer pushed me on partitioning strategy and failure recovery.

Round 4: Behavioral. STAR format, pretty standard. Biggest question was about a time I had to push back on a stakeholder request. They care about async communication skills, probably because the whole company is remote and async-first.

Comp for my offer: L3 equivalent, US remote, came in around $165k base with equity on top. Refreshes are reasonable from what I can tell.

Overall: the technical bar is real but not FAANG-tier hazing. They care that you understand data quality and can build reliable pipelines, not that you memorized graph algorithms. If you know dbt and have done serious Airflow or similar work, you'll be fine.

Happy to answer questions.

6 replies

ds_dmitri

Really helpful, thanks. Did the SQL questions involve any recursion or were they mostly aggregation/window function territory?

de_derek

No recursion. All aggregation, window functions, and one question about using NOT EXISTS vs LEFT JOIN for a specific filtering case. Recursive CTEs came up in passing when I mentioned a use case but they didn't ask me to write one.

analyst_ana

The CI/CD schema design question is clever. Makes sense they'd use domain-relevant examples instead of the usual 'design a parking lot' stuff.

sre_sol

The Kafka question in a data eng interview is standard now. Did they go deep on exactly-once semantics or stay high-level?

de_derek

Mid-level depth. They asked how I'd handle duplicate events downstream (idempotent writes) but didn't grill me on the Kafka transaction API specifics.

backend_bekah

The 'no take-home' thing is rare. Most DEs I know hate the 8-hour take-homes that companies pretend are '2-3 hours.' Points to GitLab.