Health Guardian AI

🌟 Inspiration

We were inspired by the growing gap between everyday food choices and long-term health outcomes.

Many people receive lab reports or health diagnoses but struggle to connect those numbers to their daily grocery purchases.
At the same time, sustainability is becoming a major concern in food consumption.

👉 We wanted to create a system that bridges medical data, nutrition, and sustainability into actionable, personalized insights.

Health Guardian AI is a personalized health and sustainability companion.

Parses medical reports (blood sugar, cholesterol, blood pressure, triglycerides) into structured data.
Scans grocery items via photos or barcodes to fetch nutrition info (sugar, sodium, fat, calories).
Compares food intake with health data to detect risky patterns, such as:
- High sugar intake + rising glucose levels
Generates real-time alerts with actionable advice:
> “Replace soda with seltzer, your sugar is trending high.”
Calculates a Health Score (0–100) and an optional Sustainability Score (0–100) for each choice.

Frontend: React Native with React Navigation and optional camera/barcode scanning (expo-barcode-scanner)
Backend: FastAPI with Pydantic models, hosted on lightweight cloud services
Database: Firebase Firestore for user profiles, reports, food logs, and alerts

AI/ML Components:

OCR variability: Parsing lab reports was inconsistent due to different formats and fonts.
Food recognition: Product detection from grocery photos was noisy; barcode fallback was critical.
Time constraints: Integrating AI, databases, and frontend in a 2-day hackathon required strict scope control.
Data mapping: Linking detected food items to nutrition databases needed a custom mini-catalog.

Built a working pipeline where uploading a lab report + grocery image produced real-time health alerts.
Designed a clear alert template with:
- Severity (Green / Yellow / Red)
- Because (numbers and evidence)
- Action (concrete swap)
Created a Health Score model that judges could easily understand, with subscores for glucose, BP, lipids, and calories.
Integrated sustainability scoring as a forward-looking feature.

How to combine multiple AI agents into a single orchestrated workflow:
Perception → Nutrition → Risk → Coach → Sustainability
The importance of explainability: users trust alerts more when they see the “why” behind them.
Hackathon success depends on scoping: focusing on three strong demo rules worked better than attempting ten.
Collaboration: assigning clear agent ownership across the team boosted speed and accountability.

More user-friendly UI screens for seeing customized past data.
Expand food detection with more robust ML models + real-time barcode scanning.
Add medication–nutrition interaction alerts (e.g., ACE inhibitors with potassium-rich foods).
Support longitudinal tracking with graphs for glucose, BP, and sodium intake trends.
Enrich sustainability scoring with carbon, water, and packaging metrics.
Explore clinician dashboards where healthcare providers can monitor patient adherence.
Turn this prototype into a real mobile app that empowers people to eat smarter for their health and the planet.