How to Get an OpenAI Referral
OpenAI gets a flood of resumes, so a referral from someone who cannot describe your work is worth almost nothing. The honest path is finding the person there who already has a reason to answer you.
Find your OpenAI path →Free · No sign-up · See results in ~10 seconds
✍️ Ready-to-send intro“Hi - we both studied at [your school]. I’d love to hear about your path to a company you are targeting before I apply…”
… plus everyone else in your network who can put in a good word.
See who can refer you in - pick your target company:
Why most OpenAI referral requests get ignored
OpenAI's internal referral form asks the employee to say how they know you and to rate their confidence in you. An employee who has never worked with you either lies, submits a weak referral that recruiters learn to discount, or does nothing. Most choose nothing, because a bad referral costs them credibility with their own recruiting team.
So the real bottleneck is not access. It is evidence. You need someone who can point at a thing you built, shipped, wrote, or debugged. That is why overlap matters more than seniority: a former teammate at a previous employer is worth ten cold requests to a Staff Engineer who has never seen your code.
Where your real OpenAI paths actually are
OpenAI hires heavily from a small set of places, which means your overlap odds are better than you think. Check these in order, from strongest evidence to weakest:
- Former coworkers. People who left your past employer for OpenAI. They already know your work, so the ask is a favor, not an audition. This is the highest hit-rate path and almost nobody checks it systematically.
- Open-source and research trails. If you have contributed to a repo an OpenAI engineer maintains, filed a useful issue, or cited their paper in something you shipped, that thread is a legitimate opener with built-in evidence.
- School alumni in the right function. A shared school gets your message read, but it is a weaker signal than shared work. Narrow it: same school AND same major or same lab beats same school alone.
- Second-degree bridges. A former manager who knows someone at OpenAI can make an intro that carries their credibility instead of yours. Ask them, not the OpenAI person.
Write the message so helping you costs 30 seconds
The most common mistake is asking for the referral in the first message. That forces a stranger to make a decision about your career before they know anything about you, and the safe answer is silence. Instead: lead with the overlap in the first line (we both worked on the payments team at Stripe, or we were both in the Berkeley RL lab), then one sentence on the specific role and why it maps to something you actually did, then one link to the artifact. Close with a 15-minute chat request, not a referral request. That is a decision they can make in half a minute.
The tedious part is the finding. Going through hundreds of OpenAI employees to figure out which ones went to your school or worked at your last company is hours of manual LinkedIn scrolling, which is exactly why most people skip it and send cold messages instead. FindWarmIntros does that part: you give it OpenAI plus your school or past employers, and it surfaces the people who actually overlap with you, ranks them by how strong the shared thread is, and drafts the opener around that thread. It is free, and the point is to get you to the message rather than to keep you searching.
What to do after they reply
On the 15-minute call, do not pitch. Ask what the team is actually working on, what the hiring bar looks like for this specific role, and what would make an application easy to say yes to. Those answers do two things: they tell you whether the role is even right, and they give the person enough context about you to write a real referral.
If they offer, make it trivial: send a three-line summary they can paste, with your resume link and one sentence on how you know each other. If they do not offer, wait a week, send them one useful thing (a note on a problem they mentioned, a relevant paper), and then ask directly. A specific ask after a real conversation converts far better than a cold one ever will.