Jessilyn Dunn, PhD
Associate Professor,
Duke University
Featured in the Session: Rising Minds: AI Innovations in Diabetes—NIDDK Early-Career Investigator Symposium
When
Sunday, June 7
at 8:00 a.m. CT
Where
La Nouvelle Orleans C (Level 2)
Ernest N. Morial Convention Center
What is your presentation about?
This presentation introduces LeMonAids, a foundation model we are building (based upon GluFormer) that integrates continuous glucose monitoring, diet, and physical activity data to identify intra-individual diabetes subphenotypes. By extending prior glucose modeling approaches to include behavioral and physiological context, the model captures the drivers of glycemic variability within individuals rather than across populations. We demonstrate how LeMonAids is aimed to predict postprandial glucose responses and enable personalized insights into how lifestyle factors influence glycemic management. The work highlights how multimodal, longitudinal data can be leveraged to move beyond static classifications toward dynamic, individualized diabetes phenotyping.
How do you hope your presentation will impact diabetes research or care?
This work aims to shift diabetes research toward intra-individual modeling, where variability within a person becomes the primary signal rather than noise. In practice, this could enable more precise, behaviorally actionable recommendations for glucose management, improving both self-management and clinical decision-making.
How did you become involved with this area of diabetes research or care?
My work in digital health and wearable sensing highlighted the limitations of single-modality approaches to understanding complex physiological systems like glucose regulation. This led to a focus on integrating multimodal biosignals to better capture real-world variability and enable more personalized models of health and disease.