Chennai: Bodhan AI, a Centre of Excellence in AI for Education incubated at IIT Madras, has launched four open AI models for Indian languages, opening their capabilities to edtech companies, startups, researchers, universities and government institutions looking to build multilingual education applications.
Developed in partnership with AI4Bharat, the models cover automatic speech recognition (ASR), text-to-speech (TTS), machine translation and optical character recognition (OCR). They are being released as digital public goods through open-weight models and hosted APIs on sovereign infrastructure.
The launch is positioned as an early building block of the Bharat EduAI Stack, a proposed sovereign digital public infrastructure for education that aims to support AI applications across India’s diverse linguistic landscape.
The four models have different language coverage. The ASR model supports 27 languages, OCR supports 23 languages, Bodhan-Translate supports 22 languages, while TTS supports 23 languages.
The models have been trained and optimised using NVIDIA Nemotron open models and libraries, including the NVIDIA NeMo framework for speech recognition, machine translation and OCR. Bodhan AI said its ASR model was post-trained to support Indian languages, including regional dialects and accents.
The models are also being served using NVIDIA TensorRT-LLM and vLLM inference microservices.
The wider objective is to allow organisations to build Indian-language AI applications without having to develop foundational capabilities from scratch.
For edtech companies, the models could provide APIs for integrating speech, translation and vision capabilities into existing products. Startups and technology companies can access the open-weight models and adapt them for specific applications, while researchers and universities can use them for further experimentation and development.
In education, the potential applications range from allowing students to interact with AI tutors through voice in their preferred language to enabling AI systems to read textbooks, worksheets, documents and handwritten answers.
Bodhan AI said the models will also form the technology foundation for two of its own applications, Student Tutor Bot and Teacher Assistant Bot.
Student Tutor Bot is aimed at students in Classes 6 to 12 and is designed around NCERT and SCERT curricula. Students can interact with the system through text or voice across 22 Indian languages, using textbook content to receive explanations, examples and assessments.
The platform is designed to guide students through problems step by step and includes tools for diagrams, calculations and other forms of working. Student responses can also be used to inform subsequent learning support.
The Teacher Assistant Bot, meanwhile, is designed as an AI workspace for teachers. It can generate lesson plans, worksheets, quizzes, homework and revision material based on parameters such as grade, subject, topic, duration and difficulty.
Teachers can also upload student work for evaluation against specified marking criteria.
Bodhan AI said AI-generated outputs will remain subject to teacher review, with teachers able to edit, regenerate or discard them rather than allowing the system to make classroom decisions automatically.
Prof V Kamakoti, director, IIT Madras, said the launch was a step towards building AI infrastructure that can address India’s linguistic diversity and enable wider access to AI in education.
Prof Mitesh Khapra, principal investigator at Bodhan AI, said the objective was to build a common layer that could be used by different organisations rather than having institutions duplicate foundational AI development.
The initiative also places emphasis on data governance and sovereign deployment. Bodhan AI said its architecture would incorporate data anonymisation protocols and comply with applicable national education data frameworks, while its hosted API infrastructure is designed around a sovereign deployment approach.
The models are being developed with AI4Bharat, which works on open-source datasets, tools, models and applications for Indian languages.
The launch comes as AI adoption in education increasingly moves beyond English-language applications, with multilingual speech, translation and document understanding emerging as important components for deploying AI across India’s diverse school and higher education systems.
