AI x Epistemics Research Sprint | Delhi| AI Safety India
Can AI help us think better? Or will it think for us?
The official Delhi hub for Apart Research’s AI x Epistemics Research Sprint. Hosted by AI Safety India and the Delhi AI Alignment & Safety Collective.
AI now helps people decide what to believe and what to do. It could make hard situations easier to understand. It could also concentrate decision power and make us depend on systems nobody can check. Over one weekend, teams build the benchmarks, verification tools and products that push it the right way.
How it works
The sprint runs online from 13 to 15 November. The Delhi hub runs in person from Saturday 14 November, 5:30 PM, to Sunday 15 November, 9:00 PM. You meet other builders, form a team, and work on your project together in one room.
Important: you must also sign up on the Apart sprint page and choose Delhi as your location. Without it, your project cannot be submitted.
Schedule
The detailed schedule will be shared soon.
Research tracks
Model Epistemics & Decision-Making Evals - does the model know how solid its claims are?
Trust Infrastructure - make checking claims and tracing sources cheap and hard to cheat
Epistemic Products - tools that lead to decisions people later endorse
Open Track - bigger questions, like how to measure “decision quality improved”
What you get
$2,000 in cash prizes across all tracks
Top teams get fast-tracked to the Apart Fellowship
Mentor introductions and a path to a publishable write-up
A team and a finished research project by Sunday night
Before you come
Bring a laptop and charger. No GPU needed.
Read the Resources tab on the sprint page. Each track has a reading list and example projects.
If you plan to run models, keep an API key ready.
Who should join
Engineers, researchers, students and people from any field. No AI safety background needed. Several projects need no ML skills, only careful reading and measurement. Journalists, policy analysts, forecasters and domain experts in finance, health, law or careers are especially welcome, since you know what good advice looks like.
Submissions close Monday, 16 November, 5:29 PM IST.