Bodhan AI Launches Open AI Models For Indic Languages

Bodhan AI, an IIT Madras-incubated Centre of Excellence in AI for Education, has launched a suite of open foundational AI models for Indic languages. The initiative is a collaboration with NVIDIA and AI4Bharat, and aims to accelerate the adoption of multilingual AI solutions across the Indian language AI ecosystem, with a focus on education.

The models developed by Bodhan AI and NVIDIA cover four core capabilities: speech recognition, text-to-speech, machine translation, and optical character recognition. These models have been developed using the NVIDIA NeMo framework, which includes post-training NVIDIA Nemotron 3.5 ASR to support Indian regional dialects and accents.

According to Prof Mitesh Khapra, Principal Investigator at Bodhan AI and AI4Bharat, the partnership with NVIDIA accelerates the process of bringing open, state-of-the-art capabilities to developers. The foundational models are accessible through open-weight releases and hosted APIs within India's sovereign digital ecosystem.

The initiative has significant implications for education in India, as educational applications leveraging these models will remain free for learners, teachers, and partnering state governments. This move is expected to promote the adoption of AI solutions in education and bridge the language gap in the country.

The development of these AI models is a crucial step towards promoting digital inclusion and accessibility in India. With the support of NVIDIA and AI4Bharat, Bodhan AI aims to create a robust ecosystem for Indic languages, enabling developers to build innovative applications that cater to the diverse linguistic needs of the country.

The use of NVIDIA TensorRT LLM and vLLM microservices for inference ensures that the models are efficient and scalable, making them suitable for a wide range of applications. The open-weight releases and hosted APIs will enable developers to integrate these models into their applications, promoting the development of multilingual AI solutions.

The launch of these AI models is a significant milestone in the development of India's language AI ecosystem. With the focus on education, this initiative has the potential to transform the way people learn and interact with technology in India.

In conclusion, the collaboration between Bodhan AI, IIT Madras, and NVIDIA has resulted in the development of open foundational AI models for Indic languages, covering speech, translation, text-to-speech, and optical character recognition capabilities. This initiative is expected to accelerate the adoption of multilingual AI solutions in education and promote digital inclusion in India.

The significance of this initiative lies in its potential to bridge the language gap in India and promote digital accessibility. With the support of NVIDIA and AI4Bharat, Bodhan AI is poised to create a robust ecosystem for Indic languages, enabling developers to build innovative applications that cater to the diverse linguistic needs of the country.

Overall, the launch of these AI models is a major step forward in the development of India's language AI ecosystem, and is expected to have a significant impact on education and digital inclusion in the country.