CognitiveLab, a research group based in Bangalore, has clinched the prestigious 2024 Meta Llama Impact Grant, marking a significant milestone in India’s AI innovation landscape. The grant recognizes CognitiveLab’s pioneering work on the Nayana Project, a multilingual and multimodal AI initiative designed to bridge the digital divide by expanding access to artificial intelligence across more than 22 Indic languages, with the ambitious goal of reaching over three billion people worldwide.
Empowering Low-Resource Languages with Nayana
Nayana, CognitiveLab’s flagship project, stands out as a multilingual, multimodal language model that integrates Meta’s open-source Llama models for advanced document and image processing. What sets Nayana apart is its robust support for low-resource Indic languages, offering capabilities across text, vision, and speech. The model has already surpassed existing benchmarks in optical character recognition (OCR) for ten Indian languages, addressing a critical need for inclusive AI solutions in the region.
Meta’s Commitment to Open-Source AI for Social Impact
The Llama Impact Grant program, launched by Meta in 2023, supports open-source projects leveraging Llama models-such as Llama 2, Llama 3, and Llama 4-to tackle pressing social challenges. With over one billion downloads and more than 85,000 derivative models globally, the Llama family has become a cornerstone for research and development communities worldwide.
Shivnath Thukral, Vice President and Head of Public Policy at Meta India, emphasized the transformative potential of open-source AI, particularly in a diverse country like India, stating that such initiatives are vital for bridging the digital divide and fostering innovation for billions.
Future Plans: Expanding Reach and Capabilities
With the new grant, CognitiveLab plans to:
- Broaden Nayana’s language and multimodal coverage
- Enhance its Indic tokenizer for better language processing
- Develop new datasets for speech, text, and image applications
- Release deployment tools optimized for low-resource environments
- Collaborate with the AI community to establish new benchmarks for multilingual AI
Founder Shashi Kumar highlighted that open-source tools and Meta’s Llama models have enabled CognitiveLab to build world-class AI systems like Nayana and Ambari with minimal resources, underscoring the spirit of open innovation.
A Leap Forward for Inclusive AI
CognitiveLab’s recognition by Meta signals a leap forward for inclusive, accessible AI in India and beyond. As the Nayana Project scales, it promises to empower underserved language communities and set new standards for multilingual, multimodal AI-making advanced technology truly usable for billions.