Start Here
A focused study guide for a Data Engineer interview where you'd be the company's first data hire. Read this page, then read 01 before anything technical — it changes how you answer every other question.
They are not hiring a query-writer; they are hiring the person who will own data judgement for the whole company. SQL, pipelines, and modeling are the table stakes you must clear — but the offer is decided by whether they trust you to sequence the work, build trust in the numbers, and say no to the wrong things. Demonstrate the skills and the judgement.
What this guide is
Ten chapters covering the three skill areas a first-data-hire DE loop grades — SQL, data quality & pipelines, and modeling judgement — plus the founding-hire context that frames all of them. Examples lean toward a usage-based, two-sided GPU-compute marketplace (metering, utilization, billing, supply/demand) because that's a common shape for these roles, but every pattern is general-purpose. Nothing here describes any specific company.
Two chapters are drillable: 03 — SQL Problems and 09 — Interview Q&A. Turn on drill mode to hide solutions and mark your progress; it persists in your browser.
Who this is for
- You're interviewing for a Data Engineer (or "founding data engineer" / "data platform engineer") role.
- You would be the first dedicated data person — no existing warehouse, no dbt project, maybe a few app-DB read replicas and some product analytics events.
- You're strong technically but want to make sure you're answering the founding-hire version of each question, not the big-company version.
How a typical loop maps to these chapters
First-data-hire loops vary, but they usually contain these stages. Map your prep to whichever you've been told to expect.
| Loop stage | What they grade | Chapters |
|---|---|---|
| Recruiter / hiring-manager screen | Can you frame the founding mandate? Scoping instinct. | 01, 08 |
| Live SQL | Fluency, correctness with NULLs/grain, window functions | 02, 03 |
| Pipeline / systems design | Ingestion, idempotency, schema evolution, orchestration | 04, 07 |
| Data quality / "the numbers are wrong" scenario | Debugging trust, tests, contracts, incident handling | 05 |
| Modeling whiteboard | Grain, dimensional design, SCDs, tradeoff reasoning | 06, 07 |
| "What would you do in your first 90 days?" | Sequencing, build-vs-buy, stakeholder judgement | 01, 08 |
| Behavioral / values | Autonomy, dealing with ambiguity, saying no | 09 |
The three graded skills — what "good" looks like
Not cleverness — fluency. You declare the grain of your result, write correct joins without fanning out, reach for window functions automatically, and handle NULLs deliberately. You narrate as you type. See 02 + 03.
You treat trust as the product. You design pipelines that are idempotent and re-runnable, you test at the boundaries, you know how you'd detect a silent break before a stakeholder does, and you can run an incident calmly. See 04 + 05.
You start from grain and the question being asked, not from a schema pattern. You can defend a denormalized choice and a normalized one, and you know which decisions are expensive to reverse. See 06.
Study schedule
Pick the path that matches your runway — full versions live on the hub.
- 5+ days: one section per day, drill 03 mid-week, drill 09 the day before.
- 2-3 days: 01 → 02/03 → 05 → 06 → 08 → drill 09.
- < 24 hours: 01 (framing), 03 (drill SQL), 06 (modeling), 09 (drill all). Skim 08 so you have a roadmap answer.
How to use the drills
- On drillable chapters, flip Drill mode on. Solutions collapse. Try the problem before you reveal. The reveal is worthless if you peek.
- Click practiced on each problem you've genuinely solved unaided. The counter and your progress persist locally.
- For SQL, write the answer out loud and on a timer (15-25 min each). Live coding is a performance; rehearse the performance, not just the answer.
Ready. Start with 01 — The First-Data-Hire Mandate →