Finished my HubSpot loop for a senior SWE role last month, offer accepted. The system design round was the most interesting part so figured I'd write it up.
First, HubSpot doesn't formally use L-numbers the same way Google does. Internally they have levels but you'll hear "senior," "principal," "staff" in recruiter conversations, not L5. That said, the bar for senior felt roughly equivalent to L5 at a mid-tier company.
The round itself
One hour, Zoom, whiteboard in Miro. Interviewer was a senior eng, not a panel. He gave me a classic CRM-adjacent prompt: design a system for tracking customer activity across touchpoints (email opens, form fills, page visits) at high volume. Exactly the kind of problem HubSpot actually deals with, which I liked.
Things they cared about, in order of how much time we spent: Data model first. They wanted me to talk through entities and relationships before jumping to infrastructure. I had to resist my instinct to start drawing queue diagrams. Event ingestion at scale. Kafka came up naturally. They didn't quiz me on Kafka internals but wanted to know I understood fan-out, consumer groups, ordering guarantees. Storage decisions. Write-heavy vs read-heavy tradeoffs. I talked through Postgres for relational stuff, S3 + columnar format for analytics, and they pushed on why not just one DB. API surface. How would you expose this to downstream services and third-party integrations.
Less emphasis on deep distributed systems theory (no CAP theorem quizzes). More emphasis on real tradeoffs you'd make at a real product company. HubSpot's infra is not FAANG-scale, and they know it. Design for "works reliably at a few billion events a day" not "design Twitter."
Behavioral piece embedded in design
About 20 minutes in he asked "tell me about a time you designed something at this scale for real." It wasn't a surprise, they bake behavioral into technical rounds sometimes. Have a real story ready, not a hypothetical.
Verdict
I prepped with LeetCode system design resources and found them overkill for HubSpot. The Designing Data-Intensive Applications book is more aligned with what they care about. Focus on event-driven architecture, data modeling, and being able to articulate tradeoffs out loud without getting defensive when pushed.