Track 01 Not started
AI Reasoning with Game Theory
Rigorous foundations for how multiple agents decide, cooperate and compete.
8–12 contributors
Focused Research Organization · India
IIDA Innovators & Industry Deeptech AI Alliance
Scientific advancements to market.
What we do
IIDA is a non-profit research lab that owns what it builds. We work on applied AI problems that are commercially real in three to five years — too long for a startup, too applied for a university — and turn them into IP, products and exports for India’s SMEs.
United States
Commercialises long-horizon research through strong academia, industry and capital linkages.
China
Couples research to industrial policy and manufacturing scale.
India
Comparable talent. Fewer mechanisms to convert research into IP, products and exports. That gap is what IIDA is built to close.
The gap
Why this is unowned
Nobody’s incentives point at the middle.
Deeptech pays well: category leadership, premium outcomes, real productivity gains, compounding advantage. But given the depth of the economy and per-capita income, venture capital and startups pick the low-hanging fruit over long-gestation work. Academia picks what favours publication and placement. Industry favours quick payback.
What is left in the middle is a set of high-impact problems that can be built in India, exported to the world, and deployed domestically to companies willing to move first. That white space is the whole brief.
Academic incentives disfavour large-scale teamwork across disciplines and projects with a low chance of publishable results, while the commercial profit motive precludes producing public goods. Projects between the two often go untouched.
— the case for Focused Research Organizations, US and UK
Owns, doesn’t aggregate
Not a matchmaking platform. IIDA owns the IP and the equipment, and creates the assets its research needs.
Self-funding by design
After initial capital, the target is at least 30% of revenue from industry contracts by Year 2. That’s Fraunhofer’s discipline, not a grant habit.
Built for SMEs and export
A deliberate market segment. Research that scales commercially in three to five years, aligned to IT services and product exports.
Public scoreboard · live
As of 11 August 2026 · Updated monthly · next 31 August 2026
0 / 3
Papers published 2 in progress
Year 1 target: 3 original papers
0 / 2
Open releases shipped 1 industry report in progress
Benchmarks, datasets, reports, pattern libraries. Year 1 target: 2
0
Active contributors Across five tracks
0 / 3
Institutional partners MoUs signed with universities
These are zeros, and they are dated. We publish the count before there is anything to count, because an organisation that only starts reporting once the numbers flatter it is not reporting. Full scoreboard, including discontinued tracks →
Year one · five tracks
Research
Five questions, then a focus.
Every track carries written kill criteria — the conditions under which we shut a track down — from day one. By the fourth quarter we commit to a focus, and say publicly what we stopped and why.
Every lead position is open — we’re recruiting in public rather than announcing names we haven’t secured. Two of the four are briefed but not yet started, and are marked as such rather than counted as progress.
Track 02 Seeking lead
World Models and JEPA
Predictive architectures that learn structure from far less data.
6–10 contributors
Track 03 Not started
UX and UI in Agentic Applications
The interaction layer is the adoption bottleneck, and nobody owns it.
6–10 contributors
Track 04 Seeking lead
Tokenomics — Valuing AI Token Usage
The unit economics of inference, which almost nobody can currently model.
6–10 contributors
Working on now
In progress
Not a plan. Work in hand.
The tracks above are what year one intends. This is what is actually being done this month — smaller, less tidy, and the only part of this page that can be checked today.
Since August 2026
Maharashtra AI Policy
Maharashtra is writing AI policy while the industrial base it applies to — Pune's manufacturing and services SMEs — has almost no representation in the drafting. Policy written without that input tends to fund capability that already exists in large firms and misses the constraints an SME actually faces. The window for evidence to be useful is while a document is being drafted, not after it is published.
The plan
Three years
Four exciting tracks and eager volunteers is how scope creep happens.
So the narrowing is scheduled, not hoped for. Year one explores. The Q4 gate closes three of four. Years two and three put everything behind what survived.
Honest accounting
Risk
What could go wrong, and what we did about it.
Eight things could sink this. We wrote them down before we started so that nobody has to discover them on our behalf later. The three most urgent are below.
| The risk | Why it’s real | The move against it |
|---|---|---|
| Credibility cold start | No track record yet. Funders and researchers both wait for first movers. | Anchor advisor first, entity second. Ship one visible artefact within 90 days. |
| Funding before proof | Indian funders want output first. Output needs funding. Chicken and egg. | Compute-first fundraising: easier to win, and it demonstrates resourcefulness. |
| Focus discipline | Four exciting tracks plus eager volunteers makes scope creep the default outcome. | Kill criteria written before work starts. A focus committed by Q4 of Year 1. |
Precedent
1988
India has built an industry this way before.
NASSCOM was founded in 1988 by thirty-eight companies as a not-for-profit industry body. It anchored the 1991 software-export framework, catalysed the STPI and SEZ model, and watched the industry it convened grow from roughly $100 million to more than $250 billion.
An industry alliance — not a government department — built India’s IT sector.
Get involved
Contribute
Low lift to all in.
The founding group is capped at fifteen people. The ask is one to two hours a week for six months, and the willingness to act without waiting for a mandate.
Low lift
Open doors
- Warm introductions to funders and Scientific Director candidates
- Names for the Advisory Council
- Compute credits through personal contacts
- Pro-bono legal, accounting or office space
- Critique of this charter before it hardens
Medium lift
Get your hands dirty
- 2–3 hrs a week as a pod lead or anchor
- Co-author or co-review the first papers
- Mentor two or three juniors a quarter
- Help write the first grant applications
- Domain advisor for one track
Higher lift
Go all in
- A twelve-month Research Lead role
- The Program Manager role
- Fundraising, or a founding patron commitment
- An industry sabbatical or secondment
- Co-apply for institutional grants