IIT Madras-incubated Bodhan AI, with AI4Bharat and NVIDIA, launched four AI models for Indic languages.
Indian Institute of Technology (IIT) Madras-incubated Bodhan AI – a Centre of Excellence focused on artificial intelligence (AI) for education – in partnership with AI4Bharat and NVIDIA, has introduced four key AI models designed for Indian languages. These cover speech recognition, speech generation, machine translation, and optical character recognition (OCR). The models are available as open weights and hosted application programming interfaces (APIs) on sovereign digital infrastructure, enabling startups, edtech firms, researchers and government bodies to build customised applications.
Together, these models form part of the Bharat EduAI Stack, which Bodhan AI envisions as a shared digital public infrastructure for India’s multilingual education system. The goal is to prevent duplication of effort, so institutions don’t have to rebuild the same AI tools separately for different languages. As (principal investigator at Bodhan AI) Prof. Mitesh Khapra explained, Bodhan AI wants to “build with the ecosystem, not compete with it,” offering a common technology layer for education-focused AI.
The models were trained using NVIDIA’s Nemotron open libraries, including the NeMo framework and optimised to handle Indian languages, dialects, and accents. They are deployed with NVIDIA TensorRT-LLM and vLLM microservices. Bodhan AI and NVIDIA are also collaborating on datasets and training recipes for future language models.
Potential applications include voice-enabled AI tutors and tools that can read and interpret educational material in multiple Indian languages. Bodhan AI is already launching two products based on these models: the ‘Student Tutor Bot’ and the ‘Teacher Assistant Bot’. Importantly, teachers remain in control, with AI-generated content serving only as a draft to be reviewed or edited. Prof. V. Kamakoti, director of IIT Madras, emphasised that India’s AI efforts must prioritise technology that respects its linguistic diversity.
