Nutrisimple.fit
Watched too many friends give up on nutrition tracking because existing apps were bloated with features they'd never use. So I built something that just works.
Click to view full size
Key Features
Download to your mobile device for native app-like experience
AI nutrition coach named Ada who actually understands "I had a burger and fries for lunch"
Works offline because your internet shouldn't dictate when you can log meals
Barcode scanning that actually works first try through native camera integration
Real-time streaming responses from AI assistant using Server-Sent Events
Timeline-based meal tracking that makes sense visually
Charts that tell you what matters, not everything we can calculate
Natural language input because nobody wants to search through 10,000 food items
Technical Architecture
Frontend
Backend
Infrastructure
PWA & Performance
Testing & Quality
Real-time Features
Technical Challenges
Streaming AI responses without WebSockets. SSE turned out to be perfect—simpler and just as real-time.
Time zones. Every meal logged at "8 AM" could mean 30+ different actual times. Dayjs and careful UTC handling fixed it.
Making barcode scanning work reliably. Pyzbar + proper image preprocessing got us to 99%+ accuracy.
Nutritionix returns wildly different data structures. Built a mapper system that handles anything they throw at us.
Offline-first architecture that syncs seamlessly. Service workers + IndexedDB + retry logic = invisible sync.
Results & Impact
AI nutrition coach that can execute tools on your behalf
Offline functionality that actually works—log meals on a plane, syncs when you land
Docker deployment that just works—one command from code to production