Finished the Stripe data engineering loop last month for a senior DE role. Wanted to write this up because data eng interview content for Stripe specifically is almost nonexistent and I had to piece it together from SWE and DS threads.
Short version: it's harder than I expected, more engineering-forward than most DE loops, and SQL is important but not the main event.
What the loop looked like:
Recruiter screen, technical phone screen, then a 4-round onsite.
Phone screen: SQL. They gave me a schema (something like transactions, merchants, events tables, felt like a simplified version of an internal Stripe schema) and asked me to write increasingly complex queries. I got window functions, a CTE-heavy question about attribution, and a question about identifying anomalies in event data. Medium-hard. If you're only comfortable with basic aggregations, study up.
Onsite rounds: Data modeling: Design the schema for a new feature. I got: design a data model to support merchant-level reporting on payout timing. Lots of discussion about normalization vs. denormalization for analytics. They pushed on slowly-changing dimensions, partitioning strategy, how you'd handle schema evolution without breaking downstream consumers. Pipeline design: Build a pipeline that ingests raw transaction events, transforms them, and produces a reconciliation report. Not "what tools would you use" but a deep dive on correctness guarantees. How do you handle late-arriving data? How do you know your pipeline is complete? What does a retry look like? This felt like the system design round for SWEs, just applied to data infrastructure. Coding: Python. One problem involving parsing and transforming a structured log format. Similar to the SWE coding rounds in format, a little lighter on algorithm depth but heavier on clean implementation. Values: Same values/behavioral round as the rest of the loop. Prepare real stories.
What I'd tell anyone prepping: Stripe cares about data correctness more than data volume. The questions aren't "how do you process a petabyte" they're "how do you know your numbers are right." That's the fintech orientation coming through. Every pipeline design question has an implied "what if this is wrong" thread running through it.
Tools they asked about: Spark, Airflow, dbt. Familiarity expected, deep expertise in one or two is enough.
I got to debrief but didn't get an offer. Feedback was that my pipeline design round was strong but I lost points on the data modeling schema evolution question. Fair.