India is taking a giant leap in artificial intelligence with BharatGen, a government-backed initiative focused on creating AI technology tailored for the country’s unique cultural, linguistic, and industrial landscape. Developed by the BharatGen Consortium under the India AI Mission, this project aims to build powerful language models that understand and communicate in multiple Indian languages—designed specifically to meet India’s needs.
What is BharatGen and Why Does It Matter?
BharatGen is more than just an AI project; it’s a vision to make AI truly Indian. Anchored at IIT Bombay’s Technology Innovation Hub and supported by the Department of Science and Technology, BharatGen has received the largest funding allocation under the ₹1,500 crore India AI Mission. It’s one of the first government-funded projects focused on building a Multimodal Large Language Model (LLM), which can handle speech, text, and other types of data.
Rishi Bal, Executive Vice President, and Professor Ganesh Ramakrishnan, Principal Investigator of BharatGen, explain that the initiative rests on three core values: sovereignty, Indianness, and accessibility.
- Sovereignty ensures India controls its AI technology, free from geopolitical risks, with full transparency on data and model development.
- Indianness means the AI understands India’s diverse languages, dialects, and cultural nuances to provide relevant, relatable responses.
- Accessibility focuses on making AI easy to use for everyone, including people in rural areas with low literacy, by supporting voice commands and affordable technology.
Building an AI Ecosystem for India
BharatGen is a non-profit effort that goes beyond creating AI models. The goal is to build an entire AI ecosystem that developers can access and improve upon. For example, they’ve already released AyurParam, a model focused on Ayurveda, allowing startups and innovators to build healthcare-related AI applications on top of it.
Real-Life Applications That Can Touch Millions
The potential applications of BharatGen’s AI are vast. It’s designed to serve government services, businesses, healthcare, education, finance, and more. One exciting example is a prototype system for government grievance redressal that lets citizens file complaints using natural language instead of complicated forms.
From helping hospitals manage Ayurveda treatments to providing AI support in banking and legal services, BharatGen aims to impact the daily lives of ordinary Indians.
Why a One Trillion Parameter Model?
Large Language Models are often measured by the number of “parameters” — essentially the amount of data and knowledge they can process. BharatGen is working toward models with up to one trillion parameters. Bigger models mean better understanding of complex language and the ability to work with multiple data types like speech and text simultaneously.
Smaller, faster models will also be developed using a process called “distillation,” where knowledge from larger models is transferred to smaller ones, making AI more accessible on devices with less computing power.
Languages and Beyond: Serving India’s Diversity
Currently, BharatGen is working with at least 16 Indian languages, with plans to cover over 22 languages soon. This diversity makes the AI not only useful for India but also offers solutions for other multilingual countries with similar challenges.
Government Support and Funding
Most of the funding comes from the Ministry of Electronics and Information Technology (MeitY) and the India AI Mission, with over 90% dedicated to computing power. The rest covers staffing, data management, and overheads. The team is also keen on working with state governments to unlock public datasets, especially in regional languages, to enhance the models’ training.
Timeline and Team
The current phase of BharatGen’s work is funded for one year, during which they plan to make major progress. The consortium includes some of India’s top academic institutions: IIT Bombay, IIT Madras, IIT Kanpur, IIT Hyderabad, IIT Mandi, IIIT Hyderabad, and IIM Indore, with more joining soon.
Over 75 students and 45 full-time researchers are actively building India’s future in AI, gaining hands-on experience in the process.
