Archive · 1 Sep 2026
BharatGen CEO Rishi Bal Details India’s Sovereign AI Strategy
BharatGen is an Indian team building artificial intelligence for people in India.
It recently showed a model called Param2 that has 17 billion parameters.
Instead of trying to make the biggest model in the world, the team wants to make one that is affordable and useful.
It is designed to understand Indian languages, dialects, names, and customs.
The team says it has created more than 20 trillion tokens of Indian data for training its models.
It uses different chip and data-center providers in India so it is less dependent on one supplier.
BharatGen also trains college students through an internship program.
Its leaders hope this approach can eventually help other countries build AI that reflects their own cultures.
BharatGen has unveiled Param2, a 17-billion-parameter multilingual AI model.
CEO Rishi Bal says the model prioritizes affordability, accessibility, and Indian use cases over maximum scale.
The initiative says it has developed more than 20 trillion tokens of indigenous data and is approaching three petabytes.
BharatGen focuses on 22 Indian languages, regional dialects, and cultural context for applications including healthcare, education, and governance.
The team includes more than 60 researchers and staff, over 100 interns, and has published more than 30 research papers.
- Who
- BharatGen, led by CEO Rishi Bal, is developing the AI system.
- What
- The initiative unveiled Param2, a 17-billion-parameter multilingual model, and described its sovereign-AI strategy.
- When
- Param2 was recently unveiled at the India AI Impact Summit; no exact date is given.
- Where
- BharatGen is housed at the Technology Innovation Hub at the Indian Institute of Technology Bombay in India.
- Why
- The initiative aims to provide affordable, accessible AI built around Indian languages, data, culture, and infrastructure.
- Model
- Param2, a 17-billion-parameter multilingual model
- Data
- More than 20 trillion tokens of indigenous data, approaching three petabytes
- Languages
- Focus on 22 Indian languages and hundreds of dialects
- Infrastructure
- Multiple chip and data-center providers located in India
- Research team
- More than 60 researchers, engineers, linguists, and managers
- Internship pipeline
- More than 100 college interns learning to build large language models
- Research output
- More than 30 papers published in top global journals, according to BharatGen