Pinterest · Primly Community

Pinterest data engineer interview, pipelines and SQL: what I ran into

analyst_ana · 4 replies

Sharing my Pinterest DE loop experience from earlier this year. Mid-senior level. Virtual, 4 rounds.

Round 1: SQL. Live coding. I got a multi-table join scenario involving an events table and a users table, compute retention metrics. They asked me to write a 7-day and 30-day retention query. Pretty standard but they wanted efficiency discussion afterward. Asked about how I'd handle the query at Pinterest scale (billions of event rows). I talked about partitioning by date and pre-aggregated intermediate tables.

Round 2: Data modeling / system design. This one surprised me. It was half systems design and half data modeling. Prompt was roughly: design a data pipeline for Pinterest's Pin engagement events, from ingestion to analytics table. They want you to talk about schema design, partitioning strategy, streaming vs. batch tradeoffs (Pinterest uses Kafka at scale), and handling late-arriving data. Know your Kafka basics. Know what exactly-once semantics means and when it matters.

Round 3: Coding (Python/Spark). Not super heavy on Spark internals but they did ask me to write a PySpark transformation. The prompt was clean/transform a dataset with some schema evolution issues. Know how to handle nullable columns and schema mismatches.

Round 4: Behavioral. Very similar to other loops. Drive for results stories, cross-functional influence, handling a production incident.

Comp I was quoted: $200-240k TC for DE II in SF. Probably more negotiating room than they initially indicate.

Overall DE bar is real but not unreasonable. The main thing they care about is that you've thought about scale problems before, not just moved data from A to B.

4 replies

finance_faye

The $200-240k range: was that base + equity or total? And did they give you the equity vesting schedule breakdown upfront?

de_derek

Total comp, 4-year vest with a 1-year cliff. They were pretty upfront about the RSU number and refresh policy. Standard 4-year at public companies, refresh grants after year 2 typically. Just ask directly, they'll give you the sheet.

infra_ines

Kafka 'exactly-once' being in a DE interview is hilarious to me. Half the engineers who work with Kafka day-to-day couldn't explain exactly-once semantics under pressure. But I get why they ask. It's a real footgun if you assume it and don't have it.

contractor_kai

Schema evolution handling in Spark is actually one of those things that separates people who've been in production data systems from people who've done tutorials. Good signal question.