Julia Komissarchik Spills the Tea on FDA Approved AI Models and Federated Learning

Welcome back to the HITea With Grace Podcast, where we sip something warm and uncover the real stories behind healthcare innovation. Today Grace Vinton is joined by Julia Komissarchik for a conversation about the challenges and opportunities behind FDA approved AI models in healthcare. Julia explains why many of these models stay confined to a single hospital, how broader access to high quality clinical data can improve generalization, and what is needed to build AI systems that work for diverse patient populations. The discussion explores the limits of small training datasets, the importance of secure and scalable de-identification, the role of multimodal clinical data, and what it takes to build trust between hospitals and research partners. Julia also breaks down what works and what does not work in federated learning and what safeguards are required to make it a viable approach for clinical AI. To close the episode, Julia shares what motivates her, how she overcomes challenges, her advice for women in healthcare and health IT, and of course the tea she brought to the pod today.