TLDR: IIIT Hyderabad is playing a leading role in the BharatGen initiative, a national mission backed by ₹988.6 crore from the Ministry of Electronics and Information Technology (MeitY), aimed at developing sovereign, inclusive, and multi-modal AI solutions for India. The institute’s contributions include Patram, India’s first Vision Language Foundational Model for Indic documents, and eVikrAI, a Vision Language Model for Indic e-commerce, both crucial for leveraging India’s linguistic and cultural diversity in AI.
India is rapidly advancing its vision for sovereign AI, encapsulated by the “AI for India and of India” initiative, with the International Institute of Information Technology Hyderabad (IIITH) emerging as a pivotal contributor to the BharatGen consortium. This national mission in Generative AI, bolstered by a substantial ₹988.6 crore funding from the Ministry of Electronics and Information Technology (MeitY) under the IndiaAI Mission, aims to forge inclusive, multi-modal AI solutions specifically tailored to India’s rich linguistic and cultural tapestry.
Spearheaded by IIT Bombay under the National Mission on Interdisciplinary Cyber-Physical Systems (NM-ICPS), BharatGen unites a formidable consortium of leading academic institutions. Alongside IIITH, key collaborators include IIT Mandi, IIT Kanpur, IIT Hyderabad, IIM Indore, IIT Madras, IIT Kharagpur, and IIIT Delhi, signifying a concerted national effort to establish self-reliant AI capabilities.
At the heart of IIITH’s significant contributions is its Vision Language Model (VLM) team, led by Prof. RaviKiran Sarvadevabhatla. This team is dedicated to developing foundational models and applications that cater specifically to Indic languages and contexts. One of their flagship innovations is Patram, heralded as India’s inaugural Vision Language Foundational Model designed for Indic documents. Trained on an impressive 7 billion parameters, Patram possesses the capability to interpret scanned and photographed documents, accurately answer queries, and can be fine-tuned for specialized domain applications such as medicine and law. Prof. RaviKiran emphasized the meticulous development process, stating, “We built Patram from scratch, curating high-quality data to ensure robust performance across diverse use cases.”
Complementing Patram is eVikrAI, another Vision Language Model developed by the same team, specifically engineered for Indic e-commerce. eVikrAI streamlines product cataloguing by translating and vocalizing descriptions in multiple Indian languages, thereby reducing manual input requirements and significantly enhancing accessibility for sellers nationwide. These groundbreaking innovations originate from IIITH’s Center for Visual Information Technology (CVIT), recognized as one of India’s premier laboratories for vision research. Prof. RaviKiran further highlighted the institute’s vibrant research environment, noting, “Our students and researchers are constantly pushing boundaries, presenting at top-tier forums and contributing to cutting-edge developments.” He also underscored the importance of fostering an ecosystem through strategic partnerships, particularly for domain-specific models in fields like science and law. Looking ahead, IIITH plans to further enhance Patram’s capabilities, including multi-page document processing and text grounding.
The BharatGen initiative itself is a project of immense strategic national importance, as articulated by IIITH Director Prof. Sandeep Shukla, who foresees its potential applications spanning finance, agriculture, legal services, and cybersecurity. “We’re proud to be shaping India’s AI sovereignty through our contributions,” he affirmed. Prof. C V Jawahar, Dean (R&D) and head of CVIT, described BharatGen as a “unique experiment” in building powerful AI models through collaborative academic efforts.
BharatGen’s ambitious roadmap reflects India’s determination to establish self-reliant AI capabilities, ensuring the nation’s competitiveness in the rapidly evolving global AI landscape. The initiative aims to develop a comprehensive stack of AI services, mirroring the transformative impact of Aadhaar and UPI on identity and payments in India. It champions open APIs, developer toolkits, and open-source frameworks to democratize AI access and build robust public digital infrastructure. Key objectives include establishing sovereign AI compute, indigenous methods for pushing data and models to the edge, and ensuring the diffusion of AI benefits across villages, small towns, and underrepresented communities. The initiative also seeks to empower grassroots stakeholders like farmers, teachers, and health workers as active participants and co-creators in the AI ecosystem.
Earlier this year, BharatGen launched Param-1, a bilingual Large Language Model (LLM) with 2.9 billion parameters, pretrained on 5 trillion tokens in English and Hindi. The next phase of BharatGen aims to expand coverage to all 22 scheduled Indian languages, integrating text, speech, and image modalities to ensure equitable access and inclusive innovation. Rishi Bal, Executive Vice President of BharatGen, emphasized the focus on accessibility, particularly through voice as a key modality, to benefit people with poor functional literacy.
Also Read:
- India Designates IIT-Madras as UN Artificial Intelligence Centre of Excellence
- India to Equip Bureaucrats with AI Skills for Enhanced Future Governance
The ₹988.6 crore funding makes BharatGen the largest beneficiary of MeitY’s ₹1,500 crore allocation under the IndiaAI Mission, underscoring the government’s commitment to nurturing a robust AI ecosystem. Union Minister for Electronics & IT, Ashwini Vaishnaw, announced this significant funding, highlighting the government’s focus on positioning BharatGen at the forefront of building India’s sovereign AI capabilities.
