An AI Agent-Driven Continuous Glucose Monitoring Management Platform for Proactive, Personalized, and Clinician-Collaborative Care

Continuous glucose monitoring (CGM) offers unprecedented visibility into glycemic dynamics,yet most real-world use remains retrospective and fragmented, limiting timely intervention and sustained  glycemic  control.  This  project  proposes  an  AI  agent-driven  CGM  management platform  that  transforms  raw  CGM  streams  into  predictive,  personalized,  and  clinician-collaborative  care  by  unifying  interpretable  analytics,  prediction,  knowledge-grounded guidance,  and  remote  care  workflows.  Specifically,  we  will  first  deliver  an  interactive visualization  and  analytics  module  that  fuses  CGM  with  contextual  health  data  into interpretable  dashboards  to  provide  data  visualization  and  outcome  tracking  over  time  to support decision making. Second, we will develop a CGM-based glucose forecasting engine that  leverages  high-frequency  glucose  trajectories  and  contextual  signals  to  predict dysglycemic events  and  deliver proactive personalized alerts. Third, we  will build  an LLM-enabled  clinically  grounded  coaching  agent  that  combines  guideline-based  knowledge  with individual CGM patterns and daily context to generate explainable lifestyle recommendations for diabetes management. Finally, we will establish a clinician-in-the-loop workflow through a  secure  provider  interface  supporting  longitudinal  summaries,  early  warning  and teleconsultation  to  enable  timely  clinical  intervention.  Overall,  the  project  aims  to  bridge continuous sensing and timely action, enabling effective and scalable digital diabetes care.