Just wrapped IBM's senior SWE loop for a hybrid NYC role, band 7. Wanted to write this up because there is almost nothing specific floating around about what the IBM system design interview actually looks like.
First: it is not a FAANG-style distributed systems deep-dive. The vibe is closer to 'design a working system and explain your choices like you'd explain it to a team' than 'defend every CAP theorem trade-off.' That's not a diss, it's just calibration.
What the session looked like: About 50-55 minutes. One SWE interviewer (appeared to be staff or principal level) and one who joined late. I was asked to design a document processing pipeline, the kind IBM sells to enterprise clients. Think: intake from multiple sources, OCR/classification, downstream routing, storage, auditability.
Things that mattered: They wanted explicit discussion of failure modes. What happens if the OCR service goes down. What's your retry strategy, how do you prevent reprocessing duplicates. IBM's enterprise DNA shows up: compliance, audit logs, role-based access control came up organically and they leaned in when I mentioned them. Throughput estimates. I wasn't grilled on specific numbers but I needed to at least gesture at 'this is a high-write, low-read pattern, here's roughly what that means for storage.'
Things that barely came up: Deep database internals. I mentioned Kafka once and got a nod, no follow-up. Cloud-specific services. They seemed fine with 'a managed queue' rather than demanding AWS SQS vs. Google Pub/Sub debate.
Level calibration note: I interviewed for senior (band 7 equivalent), not principal. I suspect the principal bar is meaningfully higher on the depth side. If you're targeting band 8+, I'd sharpen the distributed consensus and failure recovery answers.
Overall the system design round felt fair. It rewarded clarity over cleverness, which honestly is probably the right call for an enterprise software shop. Prepared with a lot of FAANG material and had to consciously shift register partway through.