Bodhan AI
Bodhan AI is the Centre of Excellence in AI for Education incubated at IIT Madras and supported by India's Ministry of Education. It publishes trained language, speech and document models with AI4Bharat.
Company & leadership
- Institutional home
- IIT Madras, Chennai, Tamil Nadu, India.
- Legal and public role
- IIT Madras Bodhan AI Foundation; a Section 8 nonprofit incubated by IIT Madras. The Ministry of Education established the Centre.
- Board
- The current website lists Veezhinathan Kamakoti, Balaraman Ravindran and Manu Santhanam.
- Research and operating leaders
- The website names Mitesh Khapra among its academic leaders, Karthik Mahadevan Mohanakrishnan as COO, Anoop Kunchukuttan as VP Research and Ganesh Krishnan as VP Engineering. No CEO is identified.
- Model development collaborators
- Bodhan AI and AI4Bharat jointly credit the Indic model releases.
The Centre develops AI for education
Bodhan's stated program is the Bharat EduAI Stack. It aims to support learning, practice, tutoring and assessment within public education systems. Its institutional identity is a nonprofit Centre, not a commercial language-model reseller.
The Ministry launched the Centre in February 2026
Education Minister Dharmendra Pradhan launched the Centre at the Bharat Bodhan AI Conclave in New Delhi. The event brought together work on school education, higher education and skills. IIT Madras identifies the Foundation as the body established for this national program.
AI4Bharat contributes to the model releases
The September 2026 speech release credits both teams. Bodhan's public code also documents its own training and serving pipelines for speech, translation and OCR. These models adapt upstream architectures; the release documentation names the base components rather than claiming that every part was trained from the beginning.
Downloads have specific hosting terms
The Indic Open Model License permits broad use and adaptation, but is not an Apache or MIT license. Direct hosting for third parties generally requires written approval, subject to stated exceptions. Redistribution and derivatives carry license obligations. Certain large-scale products need separate terms. Upstream component terms also apply.
API operations do not each create a new model
The API exposes speech recognition, translation, transliteration, speech generation and OCR. Transliteration is a capability of Indic Translate. The recognition endpoint uses the service name indic-transcribe; the reference does not identify it as Core or Flex. These service names are not separate trained releases.