did the Scale AI onsite about 6 weeks ago for a senior SWE role (backend/infra track). reporting data because the threads on this are thin.
format: virtual, 4 rounds back to back with 10-min breaks. total time commitment was about 5.5 hours including the gaps.
round breakdown:
coding 1 (60 min): medium-hard. graph problem with a twist. i won't say the exact problem but expect something that requires BFS/DFS with some state management on top. they extended it after i got the base case working.
coding 2 (60 min): medium. dynamic programming. less tricky than coding 1 but they spent more time asking me to analyze the solution, discuss edge cases, and suggest alternative approaches.
system design (60 min): data pipeline / task routing angle. see the other thread on this, it captures the pattern well. the key thing: they care about end-to-end thinking including monitoring and failure modes, not just the happy path.
behavioral (45 min): 3-4 questions. STAR format implicitly expected, they'll pull for specifics if you're vague. see the behavioral thread.
leveling signals i picked up: at senior they expect you to drive the design, not react to prompts they paid attention to how i handled ambiguity. i asked clarifying questions at the start of both the design and one of the coding rounds, and the interviewers responded well to that clean code matters. readable variable names, comments where logic isn't obvious
post-onsite: got a decision in 5 business days. the debrief process seems standard, hiring committee model.
comp on the offer: TC was around $310K for senior in SF, mix of base, bonus, and RSU vest over 4 years. felt competitive for the level. I've seen higher at FAANG but Scale's trajectory made it reasonable.