How to get a job at OpenAI

OpenAI has 787 roles open, more than any other lab we track, and the shape of them is not what the research reputation suggests. Their three largest teams are Applied AI Engineering, Technical Success and Forward Deployed Engineering, together about 123roles. All three are fundamentally about getting models working inside other people's companies.

The research organisation is real and famous. It is also not where most of the hiring is.

Forward deployed is the door people miss

A forward deployed engineer sits with a customer and builds the thing that makes the technology useful to them. It rewards a specific and uncommon combination: genuine engineering ability plus tolerance for ambiguity plus willingness to be in the room with people who are not engineers.

Most strong engineers self-select out of these roles because they read as less prestigious than research. That is exactly why the ratio of open roles to qualified applicants is better here, and the work is closer to what actually determines whether the company succeeds.

The Residency exists for people without the obvious CV

OpenAI runs a Residency, a six-month paid programme aimed at people with the potential to contribute to frontier research who do not arrive with a conventional AI research background. Residents may be considered for full-time roles afterwards.

It is the most credible route into a frontier lab for someone strong in an adjacent field, physics, mathematics, software engineering, who would not survive a research-scientist screen today. Published accounts describe the selection as multiple technical assessments probing coding fluency, mathematical foundations and the ability to make progress on ill-posed problems. It is advertised separately from the standard listings and is easy to miss.

Where the roles are

San Francisco accounts for 535 of 787. That is an overwhelming majority and the honest planning assumption. Singapore is a notable second cluster at 32, with 63 listed remote.

What is known about the process

OpenAI publishes very little about how it interviews, and its careers site blocks automated access, so this section is drawn from candidate accounts rather than from OpenAI. Treat it as indicative.

Those accounts consistently describe a resume screen, a recruiter call, a role-specific skills assessment and a final loop of several interviews, with four to eight weeks end to end and longer for research and senior engineering roles where additional committee review is involved. They also consistently report that trajectory and demonstrated ability weigh more than credentials.

Given how little OpenAI says publicly, be sceptical of any source claiming precise knowledge of their bar or their questions. A good deal of what circulates is written to sell interview preparation.

The practical version

• Look at the forward deployed and applied engineering roles before the research ones. There are far more of them and far fewer people competing.
• If you are strong but not from an AI research background, the Residency is the route designed for you.
• Assume San Francisco.
• Expect four to eight weeks, longer for research, and discount confident claims about their internal bar.

Team and location counts are read from OpenAI's live job board and refresh daily. OpenAI publishes little direct guidance for applicants and its site blocks automated access, so process details here come from published candidate accounts and should be treated as indicative rather than official. Their listings are at openai.com/careers.