Diabetes has become a major public health challenge due to its high prevalence and chronic nature, with many individuals managing the condition for decades. One major challenge in diabetes is the enormous heterogeneity associated with the disease, which necessitates personalized approaches to its prevention, diagnosis, treatment, and prognosis. To address this, the research field has generated a large amount of complex data, and has accumulated vast prior knowledge about the disease. These data and prior knowledge contain critical, but mostly hidden, information relevant to solving this challenge. However, major hurdles exist in integrating them and extracting predictive signals, including a lack of data science expertise in the diabetes research field and the absence of diabetes-specific data science and AI expert systems, models, and tools. This initiative proposes to address these issues by establishing a pilot funding program that leverages the emerging opportunities from recent data science and AI advances. It will recruit multidisciplinary teams that include both diabetes and data science experts, to (1) develop AI foundation models for diabetes; (2) validate the models with top research questions in diabetes heterogeneity; (3) disseminate the models and engagement the community for further development, validation and application, and; (4) develop use cases that demonstrate models’ potential in accelerating the tempo of research. The expected outcomes include the integration of new AI experts into the diabetes research workforce, the creation of AI models that the average diabetes researcher can use, and informative use cases demonstrating the models' potential.