Job search
GovAI DC vs. Seasonal Fellowships: Compare the Location, Track, and Project Output
Compare DC and UK fellowship location, project output and dated intake states with two completed fictional preparation records.
Choose a GovAI fellowship by the work you want to produce and the place you can attend. The DC program focuses on American AI governance and policy; UK seasonal fellowships distinguish Research and Applied tracks. An invitation to applicants worldwide does not remove a named program’s in-person or work-authorization requirements.
As checked on October 23, 2026, the Winter 2027 fellowships are marked closed. New Summer 2027 pages provide preparation leads, with different levels of date clarity. This guide reviews public information, not an application we submitted or a fellowship we attended. You can also browse recent research and engineering roles on LandOffer.ai while deciding whether a fixed-term governance project fits your search.

Separate the program, track and application state
Start with GovAI’s opportunities page, then open the exact fellowship notice. The index distinguishes DC from UK seasonal opportunities, and Research from Applied work. It also provides general interest registration. A saved interest form, a written application and an invitation to a later selection stage are different records; none should silently become the next.
The index currently mixes closed Winter notices with new Summer pages. Its upcoming section even contains a “no upcoming opportunities” sentence alongside Summer entries. Use the detailed program notice to check the named round, and preserve inconsistencies instead of treating every card as an open vacancy. Record the named cohort’s status rather than copying “Open Positions” from the index into your application log.
Make three decisions independently. First, can you meet the location and participation conditions? Second, is your preferred contribution research or another kind of project? Third, is the current round accepting applications, merely announced, or closed? Record a relocation conflict and a closed round as separate blockers, even when the project fits your background.
DC means a US policy setting and a physical location
The DC Winter 2027 notice describes a bipartisan program running January 18 to April 9, 2027, in Washington, DC. It involves supervised research with an output such as a report or policy-facing article. The notice requires US work rights and says the fellowship does not sponsor visas. It is intended to be full-time and in person, with possible part-time arrangements for certain DC-based professionals. That possibility is not general permission to participate remotely.
The official notice marks this Winter round closed. Use it to understand the route, not to initiate an application for the expired intake. A US address, citizenship assumption or technical résumé is not a substitute for checking the actual authorization and participation rules. We do not determine any reader’s legal eligibility or obtain an accommodation through this review.
For a software engineer, the substantive question is whether technical knowledge can inform a policy decision. An evaluation harness may be relevant evidence, but the fellowship proposal needs to explain its policy audience, uncertainty and practical significance. A technically detailed project with no governance question may fit another research or engineering opportunity better. For example, comparing benchmark results can reveal measurement limits, but choosing a policy response also involves judgments about acceptable uncertainty, affected groups and implementation. State where your technical evidence ends and where a policy judgment begins. This makes a work sample easier to critique across different professional and political backgrounds.
Research and Applied tracks differ in the intended contribution
The closed UK Winter Research notice describes a self-chosen research project with a supervisor. Fellows initially explore the governance landscape before settling on a proposal. Potential outputs include reports, journal articles and policy-oriented writing. The stated London appointment is full-time and in person; a remote work test in the selection process is not evidence of a remote fellowship.
The closed UK Winter Applied notice describes projects pitched by GovAI or partner organizations outside traditional research. Examples include policy engagement, events, organizational workflows and internal products. A software background can support that work, but “Applied” does not mean a conventional software development internship or a guaranteed coding assignment.
Do not sort tracks by the file extension of the final artifact. A memo can present original research or support delivery of a policy project. A tool can be a research instrument or an operational product. Ask what decision the work supports, who defines the project and what useful completion would mean. Choose a work sample that demonstrates the contribution you want to make, so a research proposal and a delivery record do not become interchangeable.
Summer 2027 is a new check, not a reopened Winter round
The UK Summer 2027 page advertises London participation from June 7 to August 27 and says applications will open in the week commencing October 26, 2026. October 23 precedes that week. Its October-opening banner therefore should not be read as proof that a complete application can already be submitted. The page offers advance interest registration; we did not submit it or test an applicant portal.
The DC Summer 2027 page gives June 7 to August 27, 2027, at the top and an October-opening banner, but parts of the body still refer to Winter and contain incomplete dates or selection months that do not form a reliable current timetable. Its lower section invites interest before applications open. Treat this as a named Summer lead requiring updated application instructions; do not manufacture a deadline from the Winter notice.
The Summer DC notice also states in-person participation and no visa sponsorship. The UK Summer notice describes temporary-work-visa sponsorship. These are program-specific descriptions, not assurance that an individual will receive authorization. Confirm the exact round’s written conditions and your own ability to attend before making financial or employment commitments.

Worked asset: choose a location and a useful output
Maya is a fictional US-based software engineer exploring AI governance. She can consider a three-month move to either city but has not obtained leave, been assessed for work authorization, applied or secured a project. Her comparison below is an editorial planning exercise, not a description of an accepted fellow.
| Decision field | DC research route | UK Research track | UK Applied track |
|---|---|---|---|
| Personal objective | Explain a technical issue to a US policy audience | Investigate a governance question through structured analysis | Deliver a useful organizational or policy-support project |
| Proposed work sample | Short note on limits of model-evaluation evidence | Reproducible analysis plan with explicit alternative explanations | Small prototype for organizing evaluation evidence and recording review status |
| Useful audience | A reader making a US policy decision | A researcher or practitioner evaluating the argument | Staff who need a reliable process rather than a new research finding |
| Location state | Possible move, leave and authorization unresolved | Possible London move, leave and authorization unresolved | Same London constraint; choosing Applied does not remove it |
| Current intake record | Winter closed; Summer instructions need another check | Winter closed; Summer opening announced for October 26 week | Same Summer page, but project fit assessed separately |
| Decision today | Retain as a US-policy research lead | Retain if she wants research ownership | Prioritize for preparation if delivery is her main goal |
Maya’s completed personal output brief reads: “I propose an internal evidence register for a hypothetical governance team. Each entry records the evaluation claim, source, date, limitations and reviewer status. The first prototype contains only invented public-style examples. It keeps a claim marked ‘unreviewed’ until a human records a review, and exports a versioned summary. It makes evidence easier to inspect; it does not decide whether a model is safe.”
Her acceptance check for the prototype is equally concrete: an unfamiliar reviewer should locate a claim’s source and limitations without searching unrelated notes; changing a review status should leave a visible record; an incomplete entry should remain visibly incomplete. These are self-chosen design checks, not GovAI’s selection rubric or a claim that the program will offer this project.
This makes the track decision inspectable. Maya prefers building a usable workflow with stakeholders over owning a new research question, so she prepares an Applied-oriented account of her past delivery work. If she instead wants to test which evaluation evidence supports a policy conclusion, she would prepare a research question and method. Neither decision resolves the intake or location questions in the table.
Worked asset: turn closed notices into a dated preparation record
Jonah is a fictional DC-based policy analyst who discovers the Winter pages after applications close. He has a short technical-policy writing sample but no fellowship application or selection invitation. He wants to preserve useful preparation without recording interest as admission to a future cohort.
Here is his completed planning record, made on October 23. The later steps are planned personal actions, not actions already taken or promised program dates.
| Record point | Evidence or action | Correct state |
|---|---|---|
| October 23, Winter discovery | Saved the DC and UK Winter notices with their closed wording | Round archived as closed; no application initiated |
| October 23, Summer discovery | Saved the new Summer pages separately | New rounds identified; not copied over the Winter records |
| October 23, UK timing check | Recorded the stated October 26 opening week | Upcoming on his check date, with no invented closing date |
| October 23, DC timing check | Noted conflicting season text and incomplete timetable | Application instructions unresolved; interest invitation is not a submission route |
| Next personal preparation session | Review his own writing for claims, sources and uncertainty | Planned work sample revision, no program assessment claimed |
| During the advertised opening week | Revisit official pages and record any actual application instructions | Planned recheck; no guarantee that the page changes on a particular day |
| Only after finding a complete current call | Compare requirements, leave and participation conditions | Application decision still separate from submission and selection |
Jonah’s completed writing-sample diagnosis reads: “The draft states a policy recommendation before explaining its evidence. I will separate observed evaluation results, assumptions about deployment and the judgment connecting them. I will add one plausible alternative explanation and identify what additional evidence could change the recommendation. The revised sample will use public information only; I will not include employer-confidential material.”
He also writes one unsent clarification question for his notes: “For the named Summer 2027 DC round, which current notice gives the application deadline and selection timetable?” It identifies the missing information without claiming that anyone replied. He does not reuse a Winter deadline, submit a referral containing someone else’s details, or mark a general interest registration as an application confirmation.
His completed planning record leaves three actions available: improve a writing sample now, inspect the announced new round when instructions are available, and maintain a separate job search. A closed notice can still inform preparation; it cannot supply an exception to its own status.
Verify participation terms before committing
Once a complete current call is available, compare dates, time commitment, location, payment, travel support, authorization and the actual selection steps. Use that round’s terms. Do not transfer a prior season’s payment amount or recruitment calendar into a new season simply because the program names resemble each other.
Keep preparation, application and offer records separate. A draft project brief is your evidence of thought; a portal confirmation documents only submission; a selection message concerns the next stage stated in that message. Treat career support and alumni examples as context; a later job requires its own actual opportunity and selection decision. If your main need is an ongoing engineering role, compare recent jobs on LandOffer.ai alongside the fellowship options rather than treating a three-month project as a promised permanent position.


