Interview Prep Hub

First Data Hire — Data Engineer Interview Prep

Prep for a Data Engineer loop where you would be the company's first data hire. Three graded skills — SQL, data quality & pipelines, and modeling judgement — wrapped in the one thing that actually decides the role: can you stand up a data function from zero, alone, with good judgement? Examples lean toward a usage-based / GPU-compute marketplace domain, kept generic.

If you only have an hour

Read 01 — The First-Data-Hire Mandate to reset your framing, drill 03 — SQL Problems with drill mode on, and skim 06 — Modeling Judgement. Then drill 09 — Interview Q&A.

Drill: Interview Q&A ~35 questions · answers hidden · progress saved Drill: SQL Problems (2 sets) 24 problems · basics + advanced The Mandate what "first hire" really tests
Section A · Orient
00 Start Here Master index, how the three graded skills map to the loop, and a study schedule. Read first. 01 The First-Data-Hire Mandate What changes when you're employee-zero for data: judgement over trivia, sequencing over completeness, trust over dashboards. The lens every other chapter is read through. Critical
Section B · SQL
02 SQL Deep Dive Grain, joins & fan-out, NULL traps, window functions, gaps-and-islands, sessionization, and what to say about performance. The most-tested live skill. Critical 02b SQL Advanced Frames (ROWS/RANGE/GROUPS), LATERAL & as-of joins, recursive CTEs, set ops & data diffing, ROLLUP/CUBE, JSON/arrays, EXPLAIN & join algorithms, skew, approximate aggregates. Critical 03 SQL Problems Worked Out 12 problems with multiple approaches and a marketplace flavor. Drill mode hides solutions; progress saved locally. Drill mode 03b SQL Problems II — Advanced 12 harder problems: as-of joins, interval merging, rolling distinct counts, interruption analytics, z-score anomalies, gap-fill/LOCF, clearing-price discovery. Drill mode
Section C · Pipelines & Quality
04 Data Pipelines ELT vs ETL, batch vs streaming, idempotency, incremental loads, backfills, late-arriving data, schema evolution, orchestration. How to design a pipeline out loud. 04b Pipelines Advanced Delivery semantics, windowing & watermarks, exactly-once, CDC internals, file/table formats (Parquet, Iceberg/Delta), orchestration patterns, DLQ, write-audit-publish, cost tuning. 05 Data Quality & Trust The dimensions of quality, testing layers, data contracts, freshness & volume SLAs, anomaly detection, and how to run an incident. Trust is the deliverable. Critical 05b Data Quality Advanced Distribution drift (PSI/KS), anomaly approaches, reconciliation patterns, CI/CD for data, unit testing transforms, data diff, SLOs/SLIs & data downtime, alert design, RCA, governance. Critical
Section D · Modeling Judgement
06 Modeling Judgement Grain first, dimensional modeling, SCD types, normalization tradeoffs, OBT vs star, event/state modeling, and the judgement calls that have no single right answer. Critical 06b Modeling Advanced Fact additivity, bridge tables & many-to-many, factless/junk/role-playing dims, late-arriving data, surrogate keys, SCD 4/6, Data Vault, activity schema, semantic layer, two-sided-market modeling. Critical
Section E · Domain & Execution
07 Marketplace / Usage-Data Domain What the data looks like at a usage-based, two-sided GPU-compute marketplace: metering, utilization, billing reconciliation, supply/demand & price discovery, unit economics. Kept general — no company specifics. 08 Stack Decisions & 30/60/90 Build-vs-buy as employee-zero, warehouse & ingestion choices, what to build first, and a defensible first-90-days plan you can present in the loop. 09 Interview Q&A + Day-Of ~35 questions across all areas plus behavioral and scoping. Drill mode hides answers. Ends with day-of tactics and questions to ask them. Drill mode
Study Paths
If you have 5+ days
  • Day 1: 00, 01 (mandate) + 02 (SQL deep dive).
  • Day 2: 02b (SQL advanced) + drill 03 (problems).
  • Day 3: Drill 03b (advanced problems) on a timer until automatic.
  • Day 4: 04 + 04b (pipelines) — design one out loud.
  • Day 5: 05 + 05b (quality) + the "number is wrong" script.
  • Day 6: 06 + 06b (modeling) + 07 (domain).
  • Day 7: 08 (rehearse 30/60/90), drill 09, re-read 01 + day-of. Sleep.
If you have 2-3 days

01, 02, 03 (drill), 05, 06, 08, then drill 09. Skim 04 and 07 headings.

If you have < 24 hours

01 (framing is the highest-leverage hour), 03 (drill the SQL — most likely to be live-coded), 06 (modeling judgement), 09 (drill all Qs + day-of). Skim 08 to have a roadmap answer ready.