Buddygo

AI coach + accountability buddy + gamified city map. Gemini runs every workflow. Designed by a doctor. Built for the 800M people with metabolic disease.

Comment

Inspiration

I am a specialist in Internal Medicine and Rheumatology and an emergency physician in Friuli Venezia Giulia, Italy. Every day I discharge patients with perfect prescriptions for type 2 diabetes, hypertension, obesity, and dyslipidemia. The prescription is the easy part. What happens after they leave the clinic is where people fail: the daily walk, the dietary habit, the discipline. No healthcare system offers meaningful support there. After years of watching patients fail not at treatment compliance but at behavioral change, I decided to build the tool I always wished I could prescribe.

What it does

BuddyGo is a wellness companion powered by Gemini 2.5 Pro that pairs users with an accountability buddy and gamifies their neighborhood map to drive consistent physical activity and habit formation. The Gemini agent generates a personalized weekly plan on sign-up, creates a new mission every Monday tied to a real zone on the user's city map, analyzes the weekly check-in every Sunday and adjusts the next week's targets, detects abandonment risk through behavioral pattern recognition, and activates the buddy system when a user goes silent. No human operator touches any workflow. Gemini runs the business.

How we built it

BuddyGo is built on Google's full AI stack: Gemini 2.5 Pro via Google AI Studio, Firebase Firestore and Cloud Functions for the backend, Google Maps API with hexagonal zone overlays, and a React PWA for the frontend. The Gemini agent has six core functions: weekly plan generation, check-in analysis, buddy matching, abandonment risk detection, calibrated intervention messaging, and wellness ping activation. The points and gamification system is designed around the clinical threshold of 8 consecutive weeks, the point where habit formation is considered established.

Challenges we ran into

The main challenge was keeping the product firmly in the wellness space rather than crossing into regulated medical device territory. In Italy the regulatory framework for digital health products is strict. Every design decision, from the language used in the app to the data we collect, was made with that boundary in mind. BuddyGo does not diagnose, does not prescribe, and does not monitor clinical parameters. It builds habits. The second challenge was designing an abandonment detection system that intervenes without creating pressure. The clinical insight is that fear-based messaging drives abandonment rather than preventing it. The intervention system uses three levels: warm encouragement, positive family impact framing, and a no-pressure wellness check. Never guilt, always care.

Accomplishments that we're proud of

The abandonment detection and intervention system is something no other wellness app has built with this level of clinical reasoning. The three-level intervention model, the wellness ping that activates the buddy when a user goes silent, and the decision to reward behavior rather than outcomes are all grounded in years of clinical experience watching what actually works with real patients. We are also proud of building a product that stays on the right side of the regulatory line without sacrificing clinical depth.

What we learned

The hardest part of building a health product is not the technology. It is knowing when not to add a feature. Every time we were tempted to add clinical monitoring, medication tracking, or diagnostic suggestions, we pulled back. The product is stronger for those decisions. We also learned that the framing of wellness goals matters as much as the goals themselves. Users respond to "walk without getting tired" more than they respond to "reduce cardiovascular risk."

What's next for BuddyGo

A clinician dashboard that allows physicians to monitor patient adherence in aggregate, not clinical data but behavioral patterns. That is where the B2B revenue model lives and where BuddyGo becomes a tool that bridges the gap between the clinic and the patient's daily life. We are also planning integration with wearables to automate walking detection, removing the self-reporting step entirely and making the check-in frictionless.

Share this project:

Updates