Data Engineer Data Manipulation (SQL/Python) Interview Questions

Data Engineer SQL rounds are a different exam from the Data Scientist version, and these 41 questions show where the line falls: what gets probed is whether a transformation stays correct on messy input, not what the retention curve says. 17 come from Meta, with TikTok, RBC Royal Bank, Bloomberg, Otter.ai and ClickHouse behind it. The shapes that recur are interval merging, collapsing overlapping bookings into one unique duration; per-group ranking to pull the highest-paid employee in each department; consecutive-week streak logic to find the customer with the most rentals in a row; snapshot diffing to work out what changed for the largest member; and raw Postgres string parsing that has to happen before any aggregation is possible. In practice that means ROW_NUMBER and RANK partitioned by department, self-joins on date offsets, LAG for streaks, COALESCE around nullable joins, and one question that asks you plainly how a pandas boolean filter differs from a SQL WHERE clause. 21 of the 41 sit in a technical screen rather than onsite, and nothing here is rated hard: 27 are medium and the remainder easy. 29 open in a live PostgreSQL console so you can run the query against the tables first, and 97% are readable without premium.