UVA Data Science researcher leads $4.7M project for AI-powered diabetes management

University of Virginia School of Data Science researcher Heman Shakeri has been awarded a major new research grant to lead work at the intersection of machine learning and diabetes care. Shakeri will serve as a contact PI alongside Dr. Greg Forlenza, pediatric endocrinologist at the University of Colorado Anschutz Medical Campus (Barbara Davis Center for Diabetes).

The award is jointly funded by Breakthrough T1D and The Leona M. and Harry B. Helmsley Charitable Trust. The project leverages a $3.9 million grant combined with $800,000 in in-kind contributions from industry partners Tandem Diabetes Care and Arecor, bringing the total project support to approximately $4.7 million.

The initiative focuses on developing next-generation, fully closed-loop insulin delivery systems. The research will test how adaptive learning algorithms, developed at UVA, combined with ultra-rapid-acting insulin formulations can personalize glucose control without requiring patients to manually intervene for meals or exercise.

University of Colorado Anschutz serves as the administrative prime, and UVA leads the algorithm development and engineering work packages. Clinical trials will be conducted at three sites: UVA, the Barbara Davis Center, and the University of California, San Francisco (UCSF).

This work is about giving people their time, attention, and peace of mind back. Type 1 Diabetes demands constant vigilance. Our goal is to design systems that learn continuously and adapt automatically so individuals with Type 1 Diabetes, especially children and their families, can live more freely."

Heman Shakeri, researcher, University of Virginia School of Data Science

A data-driven approach to a clinical challenge

Automated insulin delivery (AID) systems have transformed diabetes management, yet even advanced systems require users to announce meals and calculate carbohydrates. The Forlenza-Shakeri project addresses this by combining adaptive machine learning algorithms with ultra-rapid-acting U-500 insulin. At the core of the research is a framework developed at UVA that allows the delivery algorithm to continuously explore and refine control strategies. Rather than relying on fixed parameters, the system actively learns from patient data, adjusting to circadian rhythms, stress, and metabolic changes in real time.

This grant underscores UVA's central role in the engineering of medical therapeutics. Shakeri and the UVA Center for Diabetes Technology are responsible for the adaptive learning architecture that makes this fully automated control possible. "UVA is where the core data science and engineering innovation happens," Shakeri noted. "We are building the learning systems that allow these devices to personalize care in real time."

Building a track record of innovation

This award represents a significant milestone in Shakeri's research trajectory. It builds on previous successes, including multiple UVA LaunchPad grants and collaborative projects with the UVA Comprehensive Cancer Center. Across these efforts, Shakeri's work has adhered to a consistent theme: designing systems that respond to real human variability rather than forcing patients to adapt to rigid technology.

"Diabetes is not static," Shakeri said. "People's bodies change from hour to hour. Data science allows us to build systems that learn alongside the patient instead of working against them."

Collaboration and impact

The project is supported by a strategic collaboration between Breakthrough T1D and the Helmsley Charitable Trust.

"This exciting project builds upon a decade of collaboration between the Barbara Davis Center and UVA to continue to bring the most advanced technologies to children and adults with diabetes," said Forlenza.

With cutting-edge research infrastructure and deep clinical expertise, the UCSF Diabetes Technology Research Team will also play a key role in advancing next-generation treatment strategies for people living with diabetes through this collaboration.

As the project moves into clinical phases, Shakeri sees this work as part of a broader shift toward patient-centered AI. "This is not just about one device," he said. "It is about rethinking how we use data science to support human health in a way that is adaptive, humane, and sustainable."