Programs

Fellowships, internships, trainings.

We bring researchers, civil servants, and early-career technologists into the work. Our programs are small, well-funded, and publicly accountable.

{/* Featured fellowship */}

Flagship · Open for 2026/27

The BRAIN Fellowship. Nine months, six fellows.

A fully-funded, nine-month fellowship for early-career researchers, civil servants, and engineers working on questions of AI governance and public-sector deployment. Each fellow writes a working paper and convenes a workshop.

{[['09', 'Months'], ['06', 'Fellows'], ['01', 'Working paper each'], ['USD 18k', 'Stipend']].map(([v, l]) => (

{v}

{l}

))}

{/* Programs list */}

Five lines of work, shared between them.}/> {[ ['Fellowship', 'Nine months', 'Early-career researchers and civil servants. Stipend, housing, full research budget.', 'Open · rolling'], ['Summer internship', 'Ten weeks', 'Undergraduate interns join a team for a focused project. Paid, with mentoring.', 'Open · March'], ['Practitioner training', '2–5 days', 'Short courses for judiciary, health, and education practitioners working with ML systems.', 'Quarterly'], ['Residency', '3 months', 'A writing residency for policy writers and journalists. Desk space, stipend, library access.', 'Invite-only'], ['Teacher development', '2 days', 'Summer workshops for school teachers running AI clubs or Olympiad preparation.', 'August'], ['Civil-service exchange', '6 months', 'Secondments from the Nepali civil service to BRAIN — and the other way around.', 'Rolling'], ].map(([t, len, d, status]) => (

{len}

{status}

{t}

{d}

))}

{/* What fellows write about */}

The questions already in the room.} lede="Fellows propose their own research question. In recent years, these have clustered around four themes." />

{[ ['Public-sector ML deployments', "What does 'accountable' mean, operationally, for a triage tool in a district hospital?"], ['Meaningful consent', "What can a person in a health or welfare queue actually consent to?"], ['Evaluation standards', "How do we move from 'tested on a benchmark' to 'tested with the people affected'?"], ['Labour & displacement', "Who loses work when a ministry adopts an automated system? Who gains?"], ].map(([h, q]) => (

{h}

{q}

))}

{/* Application */}

Four steps, no gatekeeping.}/>

{[ ['01', 'Written proposal', 'A 1,000-word research proposal. We provide the template and two past-year examples. No CV at this stage.'], ['02', 'Conversation', 'Shortlisted applicants have a one-hour conversation with two members of the research team. We discuss the proposal, not credentials.'], ['03', 'Reference check', 'Two references. One can be a colleague, the other a person the proposed research would affect.'], ['04', 'Offer', 'Decisions within six weeks of the deadline. Stipend, desk, and a named mentor on day one.'], ].map(([n, t, d]) => (

{n}

{t}

{d}

))}

Convenings

Careful rooms, decidable questions.

We hold two-day convenings on specific, decidable questions — never on "AI in general". Most are by invitation, with a handful of public seats balloted each quarter.

A format, not an agenda.}/>

{[ ['Two days, twenty people', "Half a dozen affected practitioners, half a dozen ministry officials, the researchers who have written on the question, and a quiet minority from industry."], ['Chatham House, mostly', "What is said is on the record; who said it is not. A published proceedings is released within four weeks."], ['One decidable question', "We do not hold convenings on 'AI' as such. Every invitation names a specific decision in flight — a procurement rule, a standard, an operating procedure."], ].map(([h, d]) => (

{h}

{d}

))}

Every convening leaves a record.}/> setRoute('reading')}/> setRoute('reading')}/> setRoute('reading')}/>