Health schemes: IISc's CDPG builds secure AI data platform to detect frauds in Bengaluru

Bengaluru: The Centre of Data for Public Good (CDPG) at the Indian Institute of Science (IISc), Bengaluru, has developed a privacy-preserving AI platform, which recently enabled a national-level healthcare hackathon focused on faster claims processing and fraud detection under the Ayushman Bharat (AB) PM-JAY scheme.

The recently developed platform, called ‘Trusted AI Data Platform’, allowed startups, researchers and AI developers to work on real-world healthcare claims datasets without exposing sensitive patient information. 

The initiative addresses a longstanding challenge, revolving around healthcare and AI, said Rakshit Ramesh, VP (Analytics) at CDPG, talking about the process of “enabling innovation while protecting privacy”, and working within the realms of the Digital Personal Data Protection Act, 2023 (DPDPA).

According to CDPG, the platform created a secure and governed environment where participants could build, test and validate AI models for healthcare claims adjudication. Instead of sharing raw patient records, the system used controlled sandbox environments, curated datasets and blind evaluation systems that ensured sensitive data remained protected.

The platform addresses concerns revolving around AI and data protection and is scalable nationally, highlighted Ashok Krishnan, VP (commercial business) at CDPG.

Ramesh said, “The era of AI is here, and it is taking off in a big way. We live in a world where the full potential of what AI can do for humanity and for organisations is yet to be fully realised.”

One of the biggest hurdles for that is that organisations are not opening up their data fully to AI, as there are a lot of fears, especially regulatory fears. But with what our platform can do, we are able to induce a sort of trust in the system. That’s why we’re calling it the trusted AI data platform. By adding that secure layer, we are creating confidence around how AI can be used with sensitive data.”

“We are also adding the ability to audit exactly what data was shared, whether it was anonymised, whether it was de-identified, and whether consent was taken or not. All of those audit trails exist within the system,” he elaborated.

The platform’s architecture included hosted model access, sandboxed development environments and automated evaluation mechanisms designed to replicate production conditions while maintaining strict governance controls.

During the hackathon, Sunil Kumar Barnwal, CEO of NHA, said large-scale healthcare schemes require systems that are efficient, transparent and trustworthy.

He noted that CDPG’s platform provided the technological backbone for the hackathon.