went through the senior SWE system design round at UHG about two months ago. sharing the specifics because I found very little concrete info before going in.
the prompt was: "design a real-time claims adjudication system that can handle 50,000 claims per second at peak, with audit logging and HIPAA-compliant storage."
some notes on what they cared about:
throughput first. they wanted to know how I'd handle the 50k/sec number. we ended up talking about Kafka for ingestion, processing workers in a pool, and async write-back to a relational store. the interviewer pushed on backpressure handling specifically.
compliance is non-negotiable, not an afterthought. i mentioned audit logging early and that landed well. they asked follow-up questions about who has read access to audit tables, how logs get rotated, and what happens when a downstream system fails mid-claim. if you treat HIPAA as a checkbox you'll get grilled.
data residency. they asked whether I'd use a managed cloud DB and which one. when I said RDS, they wanted to know how I'd enforce data stays in a specific region. multi-region failover vs. data residency is a tension they actually deal with.
they weren't interested in clever. this is not the place to pitch a novel architecture. they want something maintainable, observable, and auditable. i made one "interesting" choice early and the interviewer kind of steered me back to boring and correct.
interviewer was senior, asked good questions, gave useful signals. the level was roughly what i'd call L5 at FAANG equivalents, maybe pushing L6 for certain roles. they use their own leveling so get clarity upfront.
one thing: bring a whiteboard or tablet if it's remote. they use their own video platform (some Microsoft Teams variant) and the built-in whiteboard was unusable. i screen-shared a Google Doc with boxes drawn and that was fine.