New Delhi: India’s healthcare odyssey will be defined not by the amount of data it generates but by how effectively that data is cleaned, curated, and applied. This was the unanimous view of experts at a panel discussion on India’s Healthcare Odyssey: Navigating the Future with AI and Next-Gen Tech.
Moderated by Prof. Anurag Agrawal, Head, Koita Center for Digital Health and Dean – BioSciences & Health Research, Ashoka University, the panel featured Sudeep Dey, CIO & CISO, Aster DM Healthcare; Dr. Tavpritesh Sethi, Head, Center of Excellence in Healthcare, IIIT Delhi; and Prasad Krishnamorthy, Head, IHX Pvt Ltd. Together, they emphasised that while India generates vast clinical datasets, the real challenge lies in curation, interoperability, and intelligent use.
Prof. Agrawal opened the discussion by highlighting the paradox of predicting technology. “Talking about next-gen tech is like rowing a boat — you face backwards, looking at the past, while moving into an unseen future,” he said. “India may not be as data-rich as it should be for its size, but the odyssey ahead will be shaped by how we turn the data we do have into something usable.”
Dey underlined the abundance of data in India but noted that clinical datasets require significant cleansing and harmonization before use. “Business data is far more curated.
Delivering a business dashboard in 7–8 months is reasonable. But building a clinical data lake and producing usable insights for physicians took us nearly 18 months,” he explained. He added that healthcare data often exists in “disjointed islands”, necessitating collaboration with OEMs and tech partners. “It took us 18 months to solve just three use cases. That’s the reality today,” he said.
Dey stressed that effective data sharing requires clear goals. “We must first decide what problem we want to solve. The data universe is vast, but without use cases, it becomes overwhelming. Use cases make sharing easier, provided we anonymize and involve the right stakeholders.”
Krishnamorthy highlighted the massive scale of India’s challenge. “If a premier institute takes 18 months to curate its data, imagine the plight of the remaining 90% of providers,” he said.
He laid out the numbers: “We’re talking about at least 45,000 providers, over 100,000 pathology labs, and countless clinics. The ABDM vision is that all this becomes structured, labeled, and analyzable. Only then can we deliver preventive care and universal coverage.”
Sharing a personal example, he recounted his mother’s care journey. “Even in urban tertiary hospitals, I had to carry bundles of files. Specialists often missed crucial details. Only because I was present did I catch these errors. A proper digital health exchange would prevent this.”
At IHX, he said, “40 per cent of cashless claims already flow through our exchange. Even though this isn’t complete clinical data, it provides valuable hospitalization information.
The next step is digitizing unstructured clinical data and creating longitudinal health records.” Krishnamorthy concluded that digital enablement across providers on a common infrastructure is the critical first step, requiring joint effort from technology companies and healthcare providers.
Dr. Sethi used a metaphor to describe AI’s role in healthcare. “If healthcare is a car of the future, data is the fuel, foundation models are the engine, and we still need steering and navigation.”
He elaborated that data, particularly longitudinal and multimodal, acts as fuel, while foundation models — including multimodal AI systems that integrate lab results, treatment charts, and ICU signals — serve as the engine. Steering involves human-in-the-loop oversight to ensure clinical safety, and navigation refers to emerging agentic AI systems that sequence tasks and execute them responsibly.
“We already have the components. The key will be to put them together into practical use cases,” Sethi said, citing his team’s ICU decision-making models.
On messy or unannotated data, he noted that modern AI architectures, such as contrastive learning, reduce reliance on costly manual labeling.
“At AIIMS, we are exploring how image-text pairings in histopathology can substitute manual annotation,” he said. He also cited global successes: “Graph neural networks at MIT discovered antibiotics like Helicin (2020) and Abaucin (2023) using public datasets. With strong collaboration between technology, biology, and healthcare, India can achieve similar results.”
Prof. Agrawal summarized the discussion in three stages: moving from dirty to clean data, the CIO’s harmonisation challenge, binding data to patients to create longitudinal personal health records in line with ABDM, and next-gen systems that leverage foundation models to go beyond operational efficiency into discovery, prediction, and innovation.
He highlighted breakthroughs such as TXGNN’s AI-powered drug repurposing for untreatable diseases, AI-enabled ECG prediction models that outperform traditional risk scores, and AI-driven molecule synthesis at startups like Lyla AI. “Foundation models show emergent capacities,” he noted.
“The challenge is that everyone thinks their data is gold and refuses to share, even though most breakthroughs come from public data.”
The panel agreed that India’s healthcare future hinges not just on generating more data, but on curating, sharing, and applying it responsibly.
Dey emphasised the importance of use-case-driven curation, Krishnamorthy pointed to digital enablement across providers, and Sethi highlighted AI architectures that can unlock value even from messy datasets.
Prof. Agrawal concluded, “Many systems abroad already have clean, longitudinal health records. India’s task is harder, but also more exciting. Our odyssey is about turning fragmented information into collective wisdom. That will take collaboration across technology, medicine, policy, and patients alike.”
