Just finished my Stripe loop for a senior SWE role and want to write this up while it's fresh because I couldn't find good signal when I was prepping.
The system design round at Stripe is 60 minutes. One interviewer, one problem. Mine was roughly: design a payment reconciliation system that handles high-throughput transaction ingestion, identifies mismatches between what merchants report and what Stripe's ledger shows, and surfaces discrepancies to internal ops teams.
Notice what that is: a real Stripe problem. Not "design Twitter" or "design Uber." They want you designing for the fintech domain. If you've never thought about idempotency, exactly-once delivery, distributed transactions, or ledger correctness, you will feel it.
What they actually cared about in roughly descending order:
Correctness over scale. I went straight to sharding Kafka partitions and they slowed me down. "Walk us through how you guarantee the reconciliation is correct before we talk about 10x scale." That was the signal. Get the data model right first.
Failure modes. What happens when your reconciliation job crashes halfway? What does partial state look like? How do you restart safely? This is where L5 vs L4 separates. Junior candidates describe the happy path. Staff candidates talk about what breaks.
Trade-offs stated out loud. I proposed an event-sourcing approach and they asked me to compare it against a simpler diff-table approach. They weren't fishing for a right answer. They wanted to hear my reasoning about consistency, operational complexity, and how the team would debug it a year later.
I did not make it to final offer (leveling mismatch, probably going L4), but the design round felt like the most substantive one I've had in years. No whiteboard, just shared doc in Google Meet.
One thing I'd tell anyone prepping: read Stripe's engineering blog. They have posts on their data pipeline, idempotency keys, and distributed systems. The interviewers clearly read those posts and may implicitly expect you have too.
Time-wise: 5-10 min scoping, 30-35 min design, 10-15 min deep dive on one component, last 5 for questions.