Progress in diabetes research is hindered, in part, by deficiencies in current experimental systems to accurately model human pathophysiology and/or predict clinical outcomes. Engineering human-centric platforms that more closely mimic in vivo physiology, however, requires thoughtful and informed design. Summarizing our contemporary understanding of the unique and critical features of the pancreatic islet can inform engineering design criteria. Furthermore, a broad understanding of conventional experimental practices and their current advantages and limitations ensures that new models address key gaps. Improving beyond traditional cell culture, emerging platforms are combining diabetes-relevant cells within three-dimensional niches containing dynamic matrices and controlled fluidic flow. While highly promising, islet-on-a-chip prototypes must evolve their utility, adaptability, and adoptability to ensure broad and reproducible use. Here we propose a roadmap for engineers to craft biorelevant and accessible diabetes models. Concurrently, we seek to inspire biologists to leverage such tools to ask complex and nuanced questions. The progenies of such diabetes models should ultimately enable investigators to translate ambitious research expeditions from benchtop to the clinic.