India Speaks, AI Listens: IIT Bombay’s Multilingual Tech Revolution

Key Developments

IIT Bombay is pioneering India's aggressive drive towards artificial intelligence that is appropriate for the country's distinct linguistic and cultural environment. The institute spearheads BharatGen, a cutting-edge effort that is fast developing India-focused large language models (LLMs) and datasets, with full support from the Department of Science and Technology.

Major Milestones

Release of 16 new datasets on the government's AIKosh platform for promoting innovation in Indian languages, scripts, and multimedia content.

Launch of 21 AI models, including Param 1, a bilingual foundational language model with 2.9 billion parameters—trained on 36% Indic language data, much more than global peers.

Datasets cover 11 Indian languages and include Sanskrit digitization, English-Sanskrit translation, speech recognition, solving math problems, visual question answering, and more.

BharatGen models are being built ground-up with Indian data, and not only fine-tuned from other international models.

Strategic Impact

BharatGen is a collaborative initiative of IITs at Kanpur, Mandi, Madras, Hyderabad, Indore, IIIT Hyderabad, and strategic industry partners.

The project is aligned with the India AI Mission and seeks to develop sovereign, generative AI solutions for governance, education, agriculture, and legal domains.

Project Udaan, in collaboration with IGNOU, will employ AI to translate learning material into several Indian languages, removing obstacles for millions of students across the country.

Vision for the Future

BharatGen dreams of an inclusive AI ecosystem that bridges data divides and makes India's technological advancements a true mirror of its diversity.

Nationally relevant content is expected through the public-private-academic collaborative model of the project, promoting innovation, encouraging participants, and ensuring that AI development benefits all sectors of Indian society.

Sources: Hindustan Times, India Today, IIT Bombay, TIH-IoT, ICAI