Data engineer interview questions

What the loop looks like

SQL questions

• Window functions. Running totals, rank within a group, deduplicate keeping the latest row. If you prepare one thing, prepare these.
• Self-joins and gaps-and-islands. Consecutive-day streaks, sessionising events.
• Aggregation with conditions - conditional counts, pivoting with CASE.
• Date handling - rolling windows, month-over-month change, handling time zones.
• The explain question - how would you speed this query up. Indexes, partition pruning, avoiding a full scan.

Worked set and practice notes are in SQL interview questions.

Data modelling questions

• Design a schema for an e-commerce order system, then handle returns and partial refunds.
• Star schema versus normalised: when would you denormalise, and what does it cost you?
• How do you handle slowly changing dimensions - and which type, and why?
• How would you model an event stream where events arrive late or out of order?

Pipeline design questions, where the loop is won

Expect an open prompt - ingest clickstream data and make it queryable - then follow-ups that break it. The follow-ups are the question:

• Idempotency. What happens if the job runs twice? A pipeline that cannot be safely re-run is the most common real-world failure.
• Backfills. How do you reprocess six months of history without taking the system down or double-counting?
• Late-arriving data. Events from yesterday showing up today, and what that does to yesterday's aggregates.
• Schema evolution. A producer adds a field, or changes a type. What breaks?
• Batch versus streaming. Expect to justify the choice on latency requirements, not on preference.
• Data quality. How would you know this pipeline silently produced wrong numbers?

The last one is the question that separates candidates. Say what you would monitor - row counts against expectation, null rates, freshness, referential checks - and that alerts go somewhere a person actually reads.

How depth changes with experience

• 0-2 years: SQL correctness, basic modelling, one pipeline you can describe end to end.
• 3-5 years: failure handling, cost awareness, the trade-off conversations above.
• 6+ years: platform decisions, migrations, how you got an organisation to trust its data.

Where CoPilot Interview fits

Disclosure: we build an interview assistant, so treat this as a disclosed pitch. CoPilot Interview is a native Windows and macOS app that puts suggested answers on your own screen in about two seconds, across coding, system design and behavioral rounds, with a permanent free tier that needs no credit card.

There is a role-specific page at data engineer interview help, and the general structure of a technical loop is covered in the coding interview cheat sheet.

Frequently asked questions

What questions are asked in a data engineer interview?

Four categories: SQL under time pressure (especially window functions and deduplication), data modelling including slowly changing dimensions, pipeline design with heavy follow-ups on idempotency and backfills, and a lighter Python coding round. Behavioral questions tend to focus on ownership of data quality incidents.

How much SQL do you need for a data engineering interview?

More than most candidates expect, and fluently rather than eventually. Window functions, self-joins, gaps-and-islands problems and conditional aggregation all come up regularly, and you are usually writing them live rather than at your own pace.

What is the hardest part of a data engineering interview?

The pipeline design round, because it is mostly about failure modes rather than the happy path. The follow-ups - what if the job runs twice, how do you backfill, what about late-arriving data - are the actual question, and they are hard to prepare without having been on call for a pipeline.

Do data engineers get LeetCode questions?

Usually lighter ones than software engineers. Expect practical Python rather than hard dynamic programming, with the difficulty budget spent on SQL and system design instead.