Went through the OpenAI TPM loop a few months back. My background is staff SWE transitioning to TPM, so I'll note where that context colored my experience.
Process overview: Recruiter screen, hiring manager conversation, technical screen, then a 4-hour onsite. Timeline was about 6 weeks total.
Recruiter + HM screens: Standard, mostly about background and why OpenAI. Have a clear answer to the second question. Vague mission-alignment answers don't work here. They probe.
Technical screen: This is meaningful. Not a coding interview but a technical discussion. They described a real infrastructure or cross-team coordination problem and asked how I'd approach it. What questions I'd ask, what dependencies I'd map, how I'd track progress, how I'd handle blockers at the leadership level. Know how to talk concretely about technical architecture without being asked to write code.
Onsite round 1 (program execution): Walk through a complex program you've managed. End-to-end. How you got alignment, how you handled scope changes, what went wrong. They went much deeper than I expected. Three levels deep into what I did when a dependency slipped by two months.
Onsite round 2 (technical judgment): More of the technical discussion format but higher stakes. Design-ish but TPM-flavored: how would you coordinate a major model deployment across safety, policy, infrastructure, and product teams? What's your rollout plan? What's your go/no-go criteria? Very OpenAI-specific framing.
Onsite round 3 (cross-functional leadership): Scenarios about working with research teams who don't want to commit to timelines, managing upward when the org is moving faster than the plan allows, stakeholder management under ambiguity. Every scenario felt like something that had actually happened at the company.
Values round: Same as you'll read in other posts. Real conversation, not a formality. They want TPMs who have actual opinions about responsible AI deployment, because TPMs are often in the room when those decisions get made.
Comp for TPM at senior level in SF: think L5-L6 SWE equivalent ranges. The equity question at OpenAI is always the interesting part of the package. Get clarity on vesting, cliff, and any secondary liquidity options before you decide.