Chennai, Sept 5:
Bodhan AI, the Centre of Excellence in AI for Education incubated at IIT Madras, has launched four open AI models aimed at strengthening multilingual education and enabling institutions, startups and researchers to build applications for Indian languages.
Developed in partnership with AI4Bharat, the models cover speech recognition, speech generation, machine translation and optical character recognition. They support between 22 and 27 Indian languages and are being released as Digital Public Goods through open-weight models and APIs on sovereign digital infrastructure.
The models form part of the proposed Bharat EduAI Stack, a sovereign digital public infrastructure initiative designed to create a common AI layer for India’s education ecosystem. The open approach will allow developers and institutions to adapt the models for their own applications instead of having to build language models from scratch.
IIT Madras Director V Kamakoti said India’s AI journey needed to be built on technology that understood the country’s linguistic diversity. Bodhan AI Principal Investigator Mitesh Khapra said the initiative was intended to work with the wider ecosystem and create a common layer while reducing duplication of effort.
The launch also includes a Student TutorBot and Teacher Assistant Bot. TutorBot is designed for Classes 6 to 12 and supports text and voice interactions in 22 Indian languages, using textbooks to provide explanations, practice and assessment.
The Teacher Assistant Bot is designed to help educators prepare lesson plans, worksheets, quizzes and assessments. Teachers will retain control over the AI-generated material and can review, modify or discard the outputs.
The models were trained and optimised using NVIDIA Nemotron open models and are intended to provide the technological foundation for a range of multilingual AI applications in India’s education sector.