Your goal: solve two or three medium SQL problems the way you would in a real data
science interview.
How it works (about 25 minutes)
- Each problem appears on the notepad: tables, sample rows, the task and the output columns.
- Ask clarifying questions first. Then type your query in the notepad, below the problem.
- Say "done" when you want Daniel to read it. He will ask you to walk through it, then ask
follow-ups about edge cases, performance and an extension.
- Any SQL dialect is fine (PostgreSQL, MySQL, SQLite, BigQuery, Snowflake).
- Daniel will not give hints or say whether you are right. Your written report comes after
the call.
- Already seen a problem? Say so and Daniel will swap it.
What scores well
- Restating the output (one row per what?) and asking about ties, NULLs and duplicates.
- Thinking out loud while you plan.
- Checking your query against the sample rows before saying done.
- Raising edge cases and performance before you are asked.