Kanpur: During the recently held India AI Impact Summit, the Health Minister, Jagat Prakash Nadda, officially launched the Benchmarking Open Data Platform for Health AI (BODH). This initiative, a joint venture between IIT Kanpur and the National Health Authority, was developed under a strategic MoU to create a secure consent framework and federated learning environment for health research.
Describing the platform as a cornerstone of privacy-first innovation, Prof. Manindra Agrawal, Director, IIT Kanpur, highlighted BODH as a secure, federated ecosystem where developers can train AI models on-site without ever accessing raw patient data—returning only refined model weights to ensure total privacy. This architecture effectively resolves the “AI Quality Testing Trilemma” by simultaneously delivering coverage, openness, and reliability through the strategic use of central private test sets and the ABDM framework.
Prof. Nisheeth Srivastava, CDIS, IITK, stated that the BODH platform is expected to replace Randomized Controlled Trials as the default third-party algorithmic auditing tool for health AI models, not only in India but globally. He added that this journey is just beginning.
Developed under the guidance of Prof. Nisheeth Srivastava, the platform empowers regulators to conduct low-cost, third-party testing with statistical confidence while allowing model providers to benchmark their algorithms against real-world datasets for credible validation.
By leveraging the Ayushman Bharat Digital Mission (ABDM), BODH facilitates access to a vast, diverse range of nationwide data, ensuring that the next generation of medical AI is specifically tailored to the unique healthcare needs of the Indian population.
BODH is designed to tackle the complex challenge of assessing AI quality in healthcare. It strategically resolves the “AI Quality Testing Trilemma” — the traditional trade-off between reliability, openness, and coverage.
Built on the ABDM framework, BODH offers a secure environment that enables developers to train on diverse datasets without breaching privacy laws, while allowing regulators to conduct cost-effective, third-party evaluations with strong statistical confidence.