
By Kashan Ajmeri | Last updated: 1 November 2026
Most AI internships ask you to sign a non-disclosure agreement before you have even met the team. This one is the opposite. The Allen Institute for Artificial Intelligence — Ai2 — builds its models in the open, publishes the data, releases the weights, and then invites interns to do the same. Its Open Ecosystem team, the group behind the Olmo and Molmo model families, is hiring research interns for Spring and Summer 2027, based in Seattle, for 12 weeks, with pay that is stated openly on the posting and a deadline of 8 January 2027 at 11:00pm Pacific Time.
That combination — a real research role, open publication, and pay published upfront — is rare enough to be worth a careful look. This guide covers what the internship actually is, who can apply, what it pays, and how to apply, with every fact taken from Ai2's official pages on 1 November 2026.
At a glance
| Detail | What it is |
|---|---|
| Role | Research Internship, Open Ecosystem |
| Host | Allen Institute for Artificial Intelligence (Ai2), Seattle, USA |
| Duration | 12 weeks, flexible start date |
| Start dates | Spring / Summer 2027 |
| Location | On-site in Seattle, Washington (not remote) |
| Pay | $94,320 (undergraduate), $106,110 (masters), $140,000 (PhD) — annual rates, see the pay section below |
| Deadline | 8 January 2027, 11:00pm Pacific Time |
| Visa sponsorship | Available; international candidates welcome |
| Apply | Via the official Ai2 Greenhouse application page (link below) |
| Application fee | None — the posting lists no fee, and you apply through Ai2's free Greenhouse form. No payment is asked for at any stage. |
About Ai2 and the Open Ecosystem team
Ai2 is a Seattle-based non-profit AI research institute founded in 2014 by the late Paul Allen. Its mission is building breakthrough AI to solve the world's biggest problems, and it works across large-scale open models, data, robotics and conservation. The institute employs more than 200 people and sits about 1.5 miles from the University of Washington, with close ties to the university's research community.
The Open Ecosystem team is the group building what Ai2 calls the world's most fully open AI models. That means not only open weights but open data and open training processes — the whole pipeline, released so other researchers can inspect, reproduce and build on it. Their recent releases, named on the internship posting itself, include Olmo 3, Molmo2, DR Tulu, OlmoHybrid and OlmoCore 3. If you have followed open-weight models at all in the last two years, you have probably used or read about something this team shipped.
For an intern, that openness changes the deal in a practical way. Ai2 states on the posting that it encourages open collaboration on intern projects, even with researchers at external institutions, and that it supports students freely publishing papers. In most industry labs, an intern's work disappears into an internal repository. Here, the work is meant to be seen — which is also why the application weighs publications and open-source contributions as heavily as it does.
Who can apply
The posting describes candidates as "a PhD candidate or a master/undergraduate student with a strong research background". Read that carefully, because there is a tension in the posting worth stating honestly: the internship information box welcomes master and undergraduate students with a strong research background, while the "What You'll Need" qualifications are written for someone pursuing a doctoral degree in Computer Science or a similar field, with research experience in machine learning, natural language processing, language and vision, or related areas.
In practice, that means a strong undergraduate or masters applicant is not ruled out, but they need the research record to back it up — this is not a first-exposure internship. The qualifications Ai2 lists are:
- Pursuing a doctoral degree in Computer Science or a similar field, with research experience in machine learning, natural language processing, language and vision, or related areas.
- Strong software engineering skills, especially with deep learning frameworks such as PyTorch.
Bonus qualifications — the things that make an application stand out — are a strong publication record at AI venues such as NeurIPS, ICLR, ICML, COLM, ACL or EMNLP; experience in foundation-model research, including training dynamics, scaling laws, efficiency, data curation, post-training methods such as SFT, DPO and RL, agentic training and infrastructure, and applications in scientific domains; and contributions to open-source research libraries such as PyTorch, Hugging Face Transformers, olmo-core or AllenNLP.
International candidates are explicitly welcome. Ai2's internship page states that international candidates are welcome to apply and that visa sponsorship is available, and the posting adds that Ai2 welcomes interest from applicants from outside the United States. That matters: many US research internships are effectively closed to students who need a visa, and this one is not.
What interns actually do
The posting is unusually specific about the work, and it is real research work, not a shadowing exercise. As a Research Intern on the Open Ecosystem team, you will:
- Define and lead a high-impact research project.
- Train and release leading models.
- Collaborate with and learn from team members across Ai2.
- Build open-source software for the research community.
- Author scientific papers for publication in a high-profile conference or journal.
Every intern is paired with one or more mentors from the Open Ecosystem team. The 12-week duration is fixed, but the start date is flexible, with Ai2 hiring for start dates across Spring and Summer 2027. The role is on-site: the posting states that persons in these roles are expected to work from Ai2's offices in Seattle, and the overview page adds that internships are typically on-site at the Seattle office. If you cannot relocate to Seattle for three months, this is not the right application — there is no remote option on the posting.
The pay, honestly
Ai2 publishes its intern pay on the posting, which is more transparency than most research labs offer. The figures are stated as annual rates, dependent on current degree enrolment:
- Research Intern, Undergraduate: $94,320 (annually)
- Research Intern, Masters: $106,110 (annually)
- Research Intern, PhD: $140,000 (annually)
Read those numbers for what they are: annual rates, not the amount paid for the 12 weeks. Ai2 does not state a monthly stipend or a 12-week total on the posting, so do not treat any shorter figure as official. For a sense of scale, a straight pro-rata calculation — 12 weeks out of 52 — comes to roughly $21,766 for an undergraduate, $24,487 for a masters student, and $32,308 for a PhD student over the 12 weeks. That is our arithmetic, not Ai2's stated pay, and the actual amount will depend on how Ai2 structures the internship payment. Even with that caution, the PhD rate is at the top end of what research internships pay, and the undergraduate rate is far above a typical university summer stipend.
The posting also lists benefits for team members, including medical, dental and vision cover, $125 per month for commuting or internet expenses, and $200 per month for fitness and wellbeing. Those are listed under the posting's Benefits section; the posting does not spell out separately which benefits apply to interns, so treat them as listed, not as guaranteed intern extras.
What makes a strong application
There is no separate document checklist published on the application page — you apply through the official form and follow what it asks for. What the posting does tell you is what the team values, and it is worth shaping your application around exactly that.
First, evidence of research, not just coursework. A publication at NeurIPS, ICLR, ICML, COLM, ACL or EMNLP is the clearest signal, but a strong workshop paper, a preprint with real results, or a substantial open-source contribution also speaks the team's language. Second, hands-on depth in the stack they actually use: PyTorch, and ideally experience with training or fine-tuning foundation models rather than only calling APIs. Third, alignment with open science — contributions to libraries like Hugging Face Transformers, olmo-core or AllenNLP, or a project where you released data or code publicly, will read as a genuine fit for a team whose whole identity is openness.
One honest note: if you have no research experience and no open-source work yet, this is a hard first application. The posting is written for people already doing research. That is not a reason to avoid applying if you are close — a strong masters project or a solid preprint can carry an application — but it is a reason not to treat this as a casual, five-minute submission. Put the time in, or pick a different opportunity.
How to apply
- Applying costs nothing. The posting lists no application fee, and the form is Ai2's free Greenhouse application — nobody will ask you to pay to apply.
- Open the official application page: Research Internship, Open Ecosystem — Ai2 (Greenhouse). This is the live application form for this exact role.
- Read the full posting, including the qualifications and the internship information box, before you start the form.
- Complete the application through the Greenhouse form. The posting does not publish a separate document checklist, so follow the form's own requirements.
- If you have questions about the role, the posting gives a contact address: olmo-careers@allenai.org.
- Submit before the deadline: 8 January 2027 at 11:00pm Pacific Time.
For background on Ai2's internship programme as a whole — including other teams and general information — see the official internships overview at allenai.org/internships.
Important dates
| Date | What happens |
|---|---|
| 1 November 2026 | Role listed as open on the Ai2 Greenhouse board when checked on this date (the posting states no opening date) |
| 8 January 2027, 11:00pm Pacific Time | Application deadline for Spring/Summer 2027 consideration |
| Spring / Summer 2027 | Internship start dates (flexible, 12 weeks) |
The posting gives a single apply-by date and does not publish interview stages or a notification date, so do not plan around dates that are not stated. The one date that matters is 8 January 2027. From 1 November, that is 68 days away — comfortable, but not a reason to wait. The strongest applications for roles like this are the ones submitted with time to spare, not the ones rushed in on the last evening.
Should you apply?
If you are a PhD student — or a masters or undergraduate student with a real research record — working on machine learning or NLP, and the idea of building models that are genuinely open appeals to you, this is one of the best-structured research internships open right now. The pay is published, the work is publishable, the mentorship is built in, and international applicants are welcome with visa sponsorship available.
If you have no research background yet, be honest with yourself about the odds, and use the time instead to build the record — a small open-source contribution, a reproduction of a published result, a workshop submission — that would make the next cycle's application a strong one. Either way, the deadline is 8 January 2027, and the application is on Ai2's official page.
For more paid research opportunities in AI, see the GovAI Summer Fellowship 2027 in London, which pays £14,000 for three months. The Impact AI Fellowship 2027 in Berlin (€5,000 stipend) is also on this blog, but its call is currently paused — see the notice at the top of that post. And if you are a student weighing research against earning on the side, our guide to 9 AI side hustles for students covers what actually pays without eating your study time.
Official links
- Apply on Official Website — Research Internship, Open Ecosystem, Ai2 (Greenhouse)
- Ai2 Internships overview — allenai.org/internships
All facts in this article were verified from Ai2's official pages on 1 November 2026: the Greenhouse application page for this role and the Ai2 internships overview. Pay figures are the annual rates stated on the official posting; the 12-week pro-rata figures are our own calculation and are labelled as such. If any detail changes, the official posting is the authoritative source.