BharatGen: India’s First Open-Source Multimodal AI Model | AI for Bharat

​India’s artificial intelligence (AI) landscape is witnessing a transformative development with the introduction of BharatGen, an indigenous, open-source, multimodal, multilingual AI foundation model tailored to the nation’s diverse linguistic and cultural fabric. Launched on September 30, 2024, BharatGen represents a significant stride towards AI self-reliance, aiming to reduce dependence on foreign models and address security concerns in critical applications.

BharatGen: India’s First Open-Source Multimodal AI Model | AI for Bharat​

Genesis and Objectives

The inception of BharatGen stems from the vision of Professor Ganesh Ramakrishnan of IIT Bombay, who identified the need for an AI model that encapsulates India’s vast linguistic and cultural diversity. Supported by the Department of Science and Technology under the National Mission on Interdisciplinary Cyber-Physical Systems (NM-ICPS), BharatGen is designed to be an open-source, multimodal, multilingual foundation model. The initiative seeks to mitigate reliance on foreign AI models, particularly in mission-critical sectors like defense, where security is paramount. ​en.wikipedia.org

Consortium and Collaborative Efforts

The BharatGen consortium is a testament to collaborative innovation, comprising 50–60 researchers and numerous student contributors from esteemed institutions such as IIT Bombay, IIT Kanpur, IIT Hyderabad, IIT Mandi, IIT Madras, and the International Institute of Information Technology, Hyderabad. Each institution focuses on specific tasks to develop models in voice, language, and vision, ensuring a comprehensive approach to AI development. ​en.wikipedia.org

Bharat Data Sagar: A Multilingual Repository

Recognizing the scarcity of training data for underrepresented Indian languages, BharatGen initiated the Bharat Data Sagar project—a multilingual repository dedicated to AI research. This repository captures the nuances of India’s linguistic landscape, often overlooked in international AI models, and serves as a foundational dataset for training AI models that accurately reflect India’s diverse languages and dialects. ​en.wikipedia.org

e-vikrAI: Empowering E-Commerce

In October 2024, BharatGen unveiled e-vikrAI, a vision-language model designed to streamline e-commerce for non-English speaking vendors. By automating the cataloging process, e-vikrAI allows sellers to generate product titles, descriptions, features, and pricing recommendations simply by providing product images. This innovation enhances accessibility by translating and vocalizing product descriptions in various Indian languages, thereby broadening market reach for local vendors. ​en.wikipedia.org

Current Progress and Future Outlook

As of February 2025, the framework for BharatGen’s AI model is complete, with the development team dedicating over a year and a half to the project. The initial version is anticipated to be available within the next four to ten months, marking a significant milestone in India’s AI journey. ​en.wikipedia.org

BharatGen’s commitment to integrating text, speech, and images into AI models underscores its dedication to inclusivity and robust solutions across Indian languages. By combining academic research with industry expertise through public-private partnerships, BharatGen aims to position India as a global leader in AI, impacting sectors such as agriculture, education, and healthcare.

Conclusion

BharatGen embodies India’s stride towards AI autonomy, reflecting a concerted effort to develop technology that resonates with the nation’s unique cultural and linguistic identity. Through collaborative endeavors, innovative projects like Bharat Data Sagar and e-vikrAI, and a focus on efficient, multimodal AI mode