Kami

We learned from clinical evidence that decentralizing treatment decisions and empowering patients to self-manage diabetes is one of the most important factors of therapeutic success.
We launched our community on Facebook as our target users spend most of the time there.
Link to the community
https://www.facebook.com/groups/hackingdiabetes

Interacting with patients taught us that patient support programs are human labor-intensive, but patients don't assign much value to human interactions. They prefer text-based communications over phone calls.
We built a paramedic assistant using GPT-3 that provides assistance in conceptual questions such as "what is TZD?", "what is insulin for?"

Kami is an event-based platform, largely dependent on following-up patients by dates, and biomarker thresholds. We have never done anything like it, so we first sought to see if this was something we could do.
We built an MVP in Django using Celery and Rabbit MQ and it we learned Kami was feasible.

One of our parents had an uncontrolled diabetes. We designed a follow-up program to bring his blood glucose levels back to normal. In 18 months we achieved to lower them to normal and reduced his risk of heart disease. We decided to automate the program.