went through the Goldman Sachs data engineering interview loop earlier this year for a role in the Engineering Division, NYC. four rounds total after a recruiter screen. here's the breakdown.
recruiter screen: 30 min, mostly resume walk and a few high-level questions about data modeling. they asked about my experience with "large-scale data pipelines" which means they wanted to hear specific numbers: rows per day, latency SLAs, something concrete.
round 1, coding + SQL: a medium-complexity coding problem (Python, graph traversal, nothing insane) and two SQL questions. the SQL focused on partitioning and aggregation over event logs. one question had ambiguous requirements on purpose and they wanted to see me clarify before writing anything.
round 2, system design for data: design a real-time trading data pipeline. ingest from market feeds, store for both low-latency queries and historical analysis. they pushed hard on: what happens when the feed drops, how do you guarantee exactly-once delivery, what's your approach to backfill. kafka came up immediately. they asked follow-ups about consumer group lag monitoring, which i genuinely had experience with so that worked out.
round 3, behavioral: this one surprised me, it was a full 45 min behavioral round. STAR format. lots of "tell me about a time you disagreed with a technical decision" and "how did you handle a production incident." GS puts more weight on this than i expected for an eng role.
round 4, technical deep dive: a senior IC walked through one of my past projects in serious detail. had to explain architectural choices i made 2 years ago. they were genuinely curious, not trying to trap me.
the pipeline + systems design is the make-or-break round. if you haven't designed a streaming data system end-to-end at least in theory, go do that before your loop.