AIKosh and BharatGen: Building AI as a Public Good

Imagine a small Indian AI startup wanting to build a model or application but lacking access to the large, high-quality, India-specific datasets that meaningful AI development genuinely requires, exactly the gap AIKosh, a centralised platform under the IndiaAI Mission covered elsewhere on this site, exists to close, offering over 1,200 datasets, along with models, toolkits and use cases, freely to researchers, startups and academic institutions rather than leaving every individual company to separately negotiate data access on its own.

This public-goods approach to AI infrastructure mirrors a broader Indian policy pattern worth recognising, the same logic behind the India Stack's Aadhaar and UPI digital public infrastructure, building shared, freely accessible foundational infrastructure that any company can build on top of, rather than leaving each company to build redundant infrastructure independently, now being applied specifically to AI datasets and models rather than payments and identity.

BharatGen represents this public-goods philosophy extended to model-building itself, an IIT Bombay-led academic consortium building Param2, a 17-billion-parameter multilingual foundational model specifically optimised for Indic languages, covered under the linguistic diversity challenge discussed under Sarvam AI and Krutrim elsewhere on this site, with an explicit mandate to enable AI adoption across enterprise, agriculture, healthcare, education and governance applications, a genuinely different model than venture-funded startups building proprietary, commercially licensed models.

This dual approach, government-backed public datasets and models alongside venture-funded private companies covered under the AI startup funding discussion elsewhere on this site, reflects a considered strategic bet, private capital alone might underinvest in foundational, broadly-shared AI infrastructure that benefits the entire ecosystem rather than any single company's competitive position, making public investment in shared datasets and open models a genuine complement to, rather than competitor against, India's private AI startup and hyperscaler investment boom.