We present the design, implementation, and deployment of an industry-first solution for real-time remote diabetes monitoring at large-scale. Our system delivers live insights to patients, via their mobile phones, regarding the current and future influence of their food choices and daily activities on their glucose levels. The platform, built using a distributed stream processing system, ingests real-time continuous glucose readings from sensors, insulin and meal information entered by patients within a mobile phone app, and activity levels from sensors embedded on the phone, and couples these with real-time analysis to determine the impact that the patient behavior will have on their glucose levels. We describe our design that achieves scale and system high availability through data and pipelined parallelism, decoupled system components, caches, and stateless jobs. An initial pilot of the system was deployed for use by a few hundred real diabetic patients and showed improvements in patient health - in terms of reduced periods out of range and experienced lows and highs. . A full-scale release will be deployed for several thousand patients by the end of in 2018.