OpenAI Scholars stipends open

OpenAI is again drawing attention to a small program with an outsized goal: helping people from underrepresented groups get a real foothold in deep-learning research. The program is called OpenAI Scholars. In its original form, it offered 6 to 10 stipends, mentorship, and a full-time study period for people who wanted to learn deep learning and ship an open-source project, even if they did not come through the usual PhD-to-lab pipeline (openai.com). That matters because the work of building frontier AI systems is still unusually gated. OpenAI’s 2018 Scholars announcement was explicit about the problem. The company said many existing AI fellowships and residencies required more experience or a longer commitment, which shut out people with families, jobs, or other obligations. So Scholars was designed as a remote program for people with U.S. work authorization in U.S. time zones, with optional desk space in San Francisco rather than a mandatory move (openai.com). The structure tells you what barrier OpenAI thought it was trying to lower. The details were concrete. The first cohort got $7,500 a month for three months, weekly one-on-one mentorship, a shared Slack, and $25,000 in AWS credits. Participants were expected to study deep learning full time, publish weekly updates, and finish an open-source project by the end (openai.com). This was not a scholarship in the usual sense. It was closer to a paid ramp into research work. OpenAI ran the program again for a 2019 cohort, keeping the same 6 to 10 person scale and adding a guaranteed interview for full-time work at OpenAI after completion. The company also stressed that no previous machine-learning experience was required, and pushed directly against the idea that AI research is only for people with doctorates (openai.com). That is the deeper point of the program. It was trying to widen the entrance to model-building roles before the field narrowed around elite credentials. By 2020, the format had shifted slightly but not philosophically. OpenAI opened applications for a third class of eight Scholars, expanded the program to four months, and raised the stipend to $8,500 a month. The company still described the target group the same way: people from underrepresented groups in science and engineering, including applicants without prior ML specialization but with solid software ability (openai.com). The final projects from that cohort show what OpenAI meant by “entry into research.” Scholars worked on interpretability, semantic parsing, seizure prediction from brain recordings, and other technically serious topics rather than toy coursework (openai.com). Then OpenAI folded Scholars into something closer to a hiring pathway. In 2021, the company launched OpenAI Residency and described it as an iteration of its earlier Scholars and Fellows programs. The change was significant. Residency moved away from a curriculum-style structure and toward direct collaboration on active projects, with participants hired as full-time salaried employees for six months. OpenAI said it had already made more than 20 full-time hires through its mentorship programs, accounting for one in six members of its technical staff at the time (openai.com). That history makes the current interest in “OpenAI Scholars” notable. The name still points back to a very specific model: a tiny cohort, paid time, close mentorship, and a deliberate attempt to bring in people who would otherwise be screened out before they ever touched frontier work. OpenAI’s current Residency page still carries that logic forward, saying it welcomes self-taught and non-traditional candidates and is meant to accelerate early researchers into advanced AI work (openai.com). The surprising part is not that OpenAI wants more talent. It is that one of the clearest paths it has built starts by paying newcomers to learn, one small cohort at a time.