IIT Bombay has launched its own AI company - BharatGen Technology Foundation. The company is looking to build multilingual, culturally rooted artificial-intelligence tools for the nation. This isn't a typical startup by students or professors. This company is a fully owned and run by IIT Bombay itself and its building India's own large language model similar to global AI models like ChatGPT, but built especially for our languages and our people.
On November 7, IIT Bombay officially registered the BharatGen Technology Foundation as a not-for-profit entity with the Registrar of Companies in Mumbai, headquartered on the Powai campus. The company's address is the IIT Bombay campus, clearly indicating that the institute wants to take a leading role in building India's future in AI.
BharatGen represents this groundbreaking effort by India's premier academic institutions to create a homegrown artificial intelligence (AI) ecosystem tailored to the country's diverse linguistic and cultural landscape. Unlike global AI models that often prioritise English and Western contexts, BharatGen aims to build foundational AI models that are 'by Bharat, for Bharat' ensuring sovereignty, inclusivity, and relevance to Indian users.
When was the idea for BharatGen born?
The seeds of BharatGen were sown in 2024 under the Department of Science and Technology's (DST) National Technology Mission, which provided initial funding to kickstart research into India-centric AI. The project gained momentum as a collaborative venture led by the Indian Institute of Technology Bombay (IIT Bombay). The AI company registration last month brings the agility needed to bridge academia and real-world deployment.
Professor Ganesh Ramakrishnan, who leads the foundation, emphasised the rationale, "Creating a dedicated company gives the team the freedom and agility needed to move these massive models from ‘research mode’ to real-world use." The foundation is fully owned and operated by IIT Bombay, marking a rare instance of an IIT directly spearheading a commercial AI entity.
BharatGen: Funding and government support
BharatGen has secured substantial backing, totaling Rs. 1,293 crore, positioning it as a flagship under India's broader AI ambitions. The breakdown includes:
- Rs. 235 crore from the DST, allocated in 2024 to support initial development of multilingual models.
- Rs. 1,058 crore from the Ministry of Electronics and Information Technology (MeitY) under the IndiaAI Mission, announced in late 2025 to accelerate scaling and deployment.
BharatGen: Key collaborations
BharatGen is a pan-India consortium, uniting top academic and research bodies to pool expertise. The leading institution is IIT Bombay, but core partners include IIT Madras, IIT Kanpur, IIT Hyderabad, IIT Mandi, IIT Kharagpur, IIT Delhi, and IIM Indore. The consortium also includes startups, industries, government agencies (eg: Bhashini for language tools), venture capitalists, and other IIITs for co-creation, data sharing, and AI skilling programs.
BharatGen: Objectives and key features
At its core, BharatGen seeks to develop foundational generative AI (GenAI) models fluent in over 22 Indian languages, dialects, and cultural nuances. It goes beyond translation to capture the 'Indian way' of communication - handling accents, idioms, and context-specific interactions.
Key features include:
> Multimodal Capabilities: Text understanding/generation, speech recognition/synthesis (including text-to-speech for accessibility), and document processing.
> Bharat Data Sagar: The world's largest India-centric dataset repository, encompassing text, speech (15,000+ hours annotated across 22 languages by Q4 2025), and images rooted in Indian history, philosophy, and culture.
> Specialised Applications: BharatGen will build tools like e-VikrAI (AI assistant for e-commerce sellers), Krishi Saathi (agri-bot with voice insights for farmers), and Patram - India’s first vision-language document AI model - that will be extremely beneficial to non-English speakers.
> Open-Source Focus: Subsets of models, weights, training recipes, and data will be released to democratise access, with 'distilled' (lighter) versions for developers to build apps without heavy compute resources.
The initiative emphasises data sovereignty, training exclusively on Indian datasets to avoid biases from foreign sources.
How BharatGen plans to compete with global LLMs like ChatGPT
Global large language models (LLMs) such as OpenAI's ChatGPT or Google's Gemini dominate with vast English-centric training data, often struggling with non-English nuances or cultural subtleties. BharatGen positions itself as a direct challenger by prioritising 'sovereign AI' that is authentically Indian, addressing gaps in inclusivity and relevance.
While ChatGPT supports some Indian languages superficially, BharatGen embeds deep cultural perspectives. For instace, understanding regional dialects, philosophical references from Indian knowledge systems, or context like rural-urban divides, ensuring more natural, empathetic interactions.
By releasing lightweight models, BharatGen lowers barriers for Indian startups and developers, who can innovate without massive GPU investments. This contrasts with proprietary global models' high API costs and data privacy concerns.
Trained on secure, India-sourced data, it mitigates risks of foreign dependencies, appealing to government and enterprises wary of geopolitical issues.
Tailored apps for Indian challenges (e.g., multilingual farming advice) create stickiness in high-impact areas like governance and healthcare, where global LLMs fall short.
Ultimately, BharatGen doesn't aim to outscale global giants in raw parameters but to lead in 'India-fit' AI, fostering a self-reliant ecosystem that exports Indian innovation worldwide.
2026 onward, BharatGen plans to rollout a full multimodal model, expanded benchmarks, and ecosystem-wide adoption; selective sharing of advanced models with government partners.
