Challenge Overview
Modern health data is scattered across wearables (Fitbit, Oura), blood tests, and various wellness tools. Most users, even with access to rich data streams, do not know how to translate their metrics into daily improved action. The gap: guidance that is truly personal, actionable, and adapts as you live your life.
This app solves that gap by:
- Aggregating real-time data from multiple health sources.
- Allowing users to set goals and track their biometrics holistically.
- Delivering AI-powered, context-aware recommendations for daily improvements.
- Providing smart product and supplement suggestions based on evolving needs.
- Enabling feedback and learning loops for continuous individualization.
Architecture and Agent Logic
High-Level Flow
- Onboarding & Data Link
- Users sign up, create profiles (age, weight, height, goals).
- Connect external data: Fitbit [future - Oura, bloodwork uploads, etc] and store in Flash.
- [future] All sources sync via n8n workflows to unified backend.
- Data Pipeline & Storage
- We are fitting the Fitbit data through Python Script and sending it to the webhook of n8n.
- n8n fetches/watches data from each service (e.g., Fitbit API, Oura API, OCR for bloodwork).
- Fitbit data is updated when Fitbit app is opened
- [future] Data is normalized and stored in a secure database.
- [future] A periodic poller tracks latest info (sleep, activity, stress metrics, etc.).
- AI Recommendation Engine
- Triggers occur on defined events (poor sleep, high stress, nutritional deficiency, etc.).
- User state, data, and goals are dynamically summarized as prompts for Claude (Anthropic AI).
- Claude analyzes context and responds with plain-language suggestions and justifications.
- We are using olamma model with RAG pipeline to process 1800 research papers based on nutrition and longevity and giving user suggestions based on this.
- Mobile App (React Native)
- Consumes unified data and recommendations via secure API.
- Presents a card-based feed of AI-driven actionable advice (with icons, metrics, clear CTA).
- Explanation modals (“Why am I seeing this?”) provide transparency.
- UI supports light/dark mode, is color-coded per context (e.g., blue for sleep, red for stress).
- Feedback UX allows thumbs up/down, “show me more like this”, and product rating to inform future recs.
- [future] Product Recommendation
- Backend tags products to health issues and trends (e.g., Vitamin D supplements if deficiency detected).
- Marketplace links for seamless education and potential purchase.
- [future] User Feedback Loop
- All user responses are logged.
- The AI adapts to preferences and outcomes, becoming more precise and personal over time.
Agent/Logic Summary
- Sync Agent: Uses n8n to continuously unify and update all health data streams.
- Trigger Logic: Watches for thresholds or trend changes (e.g., “Less than 6 hours of sleep” or “SpO₂ dips below 94%”).
- AI Prompt Generator: Assembles most recent data, goals, and user history into context-rich prompts for Claude.
- Recommendation Agent: Requests tailored actions, explanations, and—when relevant—curated product suggestions via Claude’s LLM.
- Feedback & Learning Agent: Tracks user responses and adapts future triggers for even higher relevance.
Why This Matters
This project empowers people to act on health data—automatically and holistically—via a seamless, AI-driven assistant, closing the loop between information and life-changing action.
Thank you for reviewing our hackathon project! We welcome your feedback and hope to help more users live their healthiest lives, every day.
Log in or sign up for Devpost to join the conversation.