Just finished the OpenAI software engineer interview process last month. Sharing the full loop because I was starving for recent signal when I started and couldn't find anything concrete.
Total timeline: about 6 weeks from first recruiter email to verbal offer.
Round 1: recruiter screen (30 min) Standard stuff. Background, motivations, logistics. She asked about my experience with large-scale distributed systems and what drew me to applied AI infrastructure. No technical content here, just vibe-check and role fit. She was straightforward about leveling: they were deciding between L4 and L5 and would calibrate based on coding and system design.
Round 2: async coding assessment (90 min, HackerRank) Two problems. One medium-hard graph traversal (something with cycles and shortest paths), one on string manipulation that was more clever than hard. I'd call the difficulty roughly LC medium-to-hard. They give you the full 90 minutes and you can use any language. I did Python.
Round 3: technical phone screen (60 min) One coding problem with a live interviewer. I got a problem that looked like a sliding window / two-pointer on arrays. The interviewer nudged me toward edge cases I missed. Clean code matters here, not just getting to an answer.
Round 4-8: full onsite (virtual, 5 sessions over 2 days) Two more coding rounds, one system design (design a rate limiter for an API serving millions of requests), one behavioral, and one called 'cross-functional' where a PM-type asked about how I'd approach shipping a feature with ambiguous requirements. That last one caught me off guard coming from pure-eng loops.
Debrief / timeline: got the verbal 8 days after the last onsite. Formal offer letter another 4 days after that.
One thing worth knowing: they explicitly said they hire for mission alignment. 'We want people who genuinely care about safe AGI, not people who want the brand.' Whether that's real or a script I can't tell you, but expect to explain your actual interest in OpenAI specifically.