Just finished the Plaid onsite loop last week for a senior SWE role (what they'd call L5-equivalent, though their leveling language is a bit looser than FAANG). Going to dump everything on the system design round here because it's the part I had the least prep for.
First: the format. 45 minutes, one interviewer, whiteboard-style in the virtual sense (they use a shared doc or occasionally Excalidraw, not a coding pad). You don't code anything. It's entirely design and tradeoffs.
The question I got was basically: design a webhook delivery system. Makes sense given what Plaid does. Think idempotency, retry logic, at-least-once vs exactly-once semantics, how you'd handle a subscriber going down for 6 hours. They really wanted me to talk through failure modes, not just the happy path. Sketch your data model, describe the queue, explain how you'd handle backpressure.
What I noticed: the interviewer pushed hard when I said things like "we'd use a message queue." They wanted specifics. Kafka vs SQS and why. What happens when a webhook consumer is timing out? Do you dead-letter? Circuit break? How do you surface that to the API customer?
Fintech-specific angle matters. Plaid's whole thing is reliability for financial connections. They're not just building CRUD; they're building infrastructure that banks and consumers depend on not failing silently. I think that context helps a lot in the design interview. Frame everything around data integrity, auditability, failure isolation.
I also got one round that blended distributed systems with API design. Things like: how would you version a public API while keeping backward compat for thousands of existing integrations? Classic fintech concern.
Leveling signal: if you're interviewing for senior/L5 at Plaid in 2026, they expect you to drive the design without much scaffolding. The interviewer was friendly but didn't volunteer hints. You need to structure it yourself: requirements, high-level design, component deep dive, failure modes, scalability.
Prep I'd recommend: Designing Data-Intensive Applications for the fundamentals, plus reading Plaid's engineering blog. They actually write about their internal systems. Helpful to know what choices they've actually made.