Listen the episode on Spotify or Apple Podcasts
Dr. Manasi Aichmueller-Ratnaparkhe sat down with Laurent Cochet on the AI InterConnect Podcast for a wide-ranging conversation on why so much of today’s health AI still learns from a narrow slice of the world’s population, and what PAICON is doing about it.
A Problem Most People Never Notice
Manasi traced her path from cancer genetics research in the US to her PhD at the German Cancer Research Center (DKFZ), where she first saw how narrow the population base behind most cancer research really was. That gap between the data used and the patient in front of the doctor became the founding question behind PAICON.
Data-First Approach
Manasi described PAICON as a data-first company before it is an AI company. PAICON has built a dataset spanning 60 countries and has expanded beyond cancer into other disease areas and data modalities. She pointed to the risks of skipping this step: models trained on one population can fail, and even hallucinate, when deployed on another, eroding clinician trust in AI altogether.
She also spoke to the double diversity problem: not just genetic and biological differences across populations, but technical differences in how data is captured, from imaging equipment quality to digitization levels across healthcare systems worldwide.
Building Toward Equitable Health AI
For Manasi, an equitable health AI model is the one trained and validated across the world’s diverse population groups, tested in real clinical settings, and honest about its own limits rather than guessing when the data isn’t there. She explained PaiX Navigator, PAICON’s platform for licensing and generating insights from its globally representative data lake, now live in beta, and PAICON’s push toward a broader, disease-agnostic data foundation built with partners across the world.
Founder Lessons and What’s Next
Asked for advice to founders building in health AI, Manasi’s answer was direct: get to market early, stay focused on one core mission instead of chasing every idea, and treat setbacks as part of the process rather than a reason to stop.
The episode also touched on her published book, remaining84: The Missing Foundation of Personalized Medicine, which explores the central problem behind PAICON’s mission: that most personalized medicine today is built on data from roughly 16% of the world’s population, leaving the remaining 84% underrepresented.
