Went through the xAI process about 6 weeks ago for an MRE position focused on pretraining infrastructure and evaluation. Thought I'd share since there's not much here yet.
Total: 4 rounds over about 3 weeks.
Recruiter screen (30 min): pretty standard, background + why xAI + availability. The recruiter was sharp, asked follow-up questions that were actually relevant. Not a checkbox exercise.
Technical screen (90 min): one interviewer, mix of ML fundamentals and a live coding problem. The coding wasn't algorithmic grind, it was more like "here's a tensor transformation, implement it cleanly and tell me why." They wanted narration. Silence was not friendly.
Research depth interview (60 min): two interviewers. They read my papers. Like, actually read them. Questions were pointed and occasionally adversarial. "Why did you make that design choice vs. X" and not letting me off the hook with vague answers. This was the hardest round.
Systems + leadership (60 min): scaling problems, how you'd approach training at scale, and some behavioral stuff about working in ambiguous environments. Not STAR-format, more conversational.
I made it to offer stage. The offer was strong but the role scope was a bit narrower than pitched, so I asked questions during the loop about autonomy and got somewhat hedged answers. Ended up declining for a different role but it was a legitimately good process. No BS rounds.