Field Building

Founded under the India AI Safety Initiative on a $65K EAIF grant. India's first Alignment Research Fellowship attracted 600+ applicants across 40 STEM universities, of whom 24 (top 4%) were selected. 10 papers from that cohort are now published or under review.

The applicant pool itself was, in my view, the most informative output — a near-complete map of where serious AI safety interest sits in Indian academic institutions, and a useful indication of where the next wave of researchers is likely to come from.

A TIA-first, mentorship-led model that surfaces technical and governance talent from regions traditionally under-served by AGISF-style fellowships.

Partner labs include Anthropic, UK AISI, FAR.AI, and others. Offers closed across 6 partners (FAR.AI, UK AISI, GovAI), with 12 accepted placements.

The premise is straightforward: instead of teaching people the field for 8 weeks and hoping they self-select into research, give them a meaningful technical problem upfront and watch which people can already do the work. My read is that this produces a better signal-to-cost ratio and better generalisation to non-EA-connected talent than the standard playbook.

A student-and-researcher initiative at IIT Delhi doing research, education, and mentorship around AI safety and alignment — modelled after HAIST and the Berkeley AI Safety Initiative, but built for the Indian subcontinent talent pool.

• Six-week AI Safety Fellowship with funding, mentorship, and access to a curated network.
• Active research on reasoning in language models, fairness benchmarks, corrigibility, and AI regulation.
• Mentorship + compute for advanced researchers, plus a technical + policy track for AGI-risk mitigation.

~12-person core team. Partners: IIT Delhi, Kairos Project.