The OpenAI Residency: How to Get In

The OpenAI Residency is a six-month, full-time, paid program that brings strong engineers, researchers, and scientists from adjacent fields into frontier AI work. It is built for people who can already ship software or prove theorems but have not worked in machine learning. It is designed to convert. Residents who clear the bar move into full-time roles as a Research Engineer, Research Scientist, or another technical member of staff. Think of it as a structured on-ramp into the lab, not a student internship.

OpenAI relaunched the residency as a six-month program in late 2025. It exists because some of the people most useful at the frontier come from physics, mathematics or pure software engineering instead of an ML PhD. The residency gives them a runway to retool, paired with a team and a mentor, and then keeps the ones who prove they can do the work. So the question behind every part of the process is not "have you done AI before" but "can you learn to do it here, fast."

The OpenAI angle

Read the residency as a hiring funnel with a training wrapper, and prepare accordingly. The mistake is to treat it as school, where showing up and being curious is enough. It is not; it is a role you interview into and then have to earn conversion from. The people who get in and stay already build or reason at a high level. They close the ML gap on purpose: implementing attention from scratch, reading model code, shipping something real in the first weeks. If you want the deep version of what that bar looks like, the Research Engineer track lesson is the closest map, because the residency selects for the same fundamentals.

Who the residency is for

The residency is aimed at a specific kind of candidate: someone with real technical depth in a field next to machine learning, who wants to cross over. In practice that means accomplished software engineers, physicists, mathematicians, and quantitative people, including career-changers and self-taught applicants who are not enrolled in a degree. This is the axis that separates it from the internship, which recruits currently enrolled students. If you have a strong engineering or research track record but no formal ML experience, the residency is usually the door built for you.

• Length: Six months, full-time. A fixed-term, salaried role, not a part-time or academic-calendar placement. You work inside a team on real problems for the duration.
• Built for: Cross-over talent. Strong engineers and scientists from fields adjacent to ML, including career-changers and self-taught applicants not enrolled in a degree. Depth somewhere real matters more than an ML resume.
• The goal: Conversion to full-time. The program is designed to hire. Residents who clear the bar move into full-time technical roles, most often Research Engineer or Research Scientist.
• The bar: The Research Engineer fundamentals. Coding fluency plus the ability to close the ML gap quickly: attention, training loops, and reading model code. Selection tests aptitude and depth, not prior ML credentials.

What selection actually tests

OpenAI publishes little about the residency loop specifically, so treat the shape below as a steer rather than a fixed script. What is consistent is the substance. The residency feeds research-adjacent roles, so its interviews lean on the Research Engineer axes. Expect strong general coding in Python, the mathematics behind modern models, and a real read on whether you can learn ML quickly. Expect a recruiter conversation, one or more technical screens, and a deeper set of rounds, with your background driving how much the bar sits on raw engineering versus research aptitude.

Get the coding section fundamentals to reflex speed: multi-head attention without the library call, and a small training loop from memory. Add the linear algebra and calculus behind backpropagation, including the softmax and cross-entropy gradients the RE loop uses as a filter. You do not need to have trained a frontier model. You do need to show you can close the gap to this work in weeks. And you need to show you already think like an engineer who ships.

The OpenAI angle

Bring a real answer on AI safety, because the residency is inside the mission, not adjacent to it. OpenAI interviews probe how you think about building powerful systems responsibly, and a program whose whole premise is bringing new people into frontier work weights that even more. The posture that lands is thoughtful, engineering-grounded concern: you take the risks seriously and you can still ship. The behavioral and values lesson covers how to prepare that conversation without sounding either like a hype merchant or someone who would freeze on the job.

Timing and how to apply

The residency runs in cohorts, and the dates and open status shift from year to year. Check the live status on OpenAI's careers page before trusting any fixed date. When a cohort is open, applications are reviewed on a rolling basis, which means an early, complete application competes against a smaller field than a late one. Apply as soon as a cohort posts, lead with the depth you already have, and be explicit about why you want to cross into AI now.

If you are a currently enrolled student rather than a career-changer, the residency is probably not your door; the internship and new-grad path is. A residency conversion becomes a normal OpenAI offer with the same moving parts. Leveling, the "Member of Technical Staff" title and offer structure are covered in the process overview and the negotiation lesson.

Common questions