Coach Calories

Context

Nutrition-tracking app designed to make daily calorie and macronutrient counting simple, without tedious data entry.

Problem

Classic nutrition apps make you weigh and search for every food in a database. Users give up before the habit sticks.

Solution

A PWA where users describe their meal in plain language. The AI automatically resolves the foods, quantities and macronutrients. Personalized daily targets replace manual counting.

My role

Product design, architecture, full-stack development, AI integration (OpenRouter), PWA, passwordless authentication and deployment.

Architecture

• TanStack Start and React for a fast PWA
• Convex for data, auth and server functions
• OpenRouter for AI meal resolution
• Tailwind v4 for the interface
• Vercel for deployment

Challenges addressed

• Turning free-form descriptions into structured foods with macronutrients
• Offering a smooth passwordless experience (OTP + Google)
• Keeping the app fast despite AI calls
• Respecting privacy: photos are temporary and never stored

Verifiable results

• Installable PWA on iOS and Android
• Automatic meal resolution from plain language
• Personalized daily targets (calories, protein, carbs, fat)
• Barcode scanning, history and trends
• Passwordless authentication (magic link + Google)

Next steps

Add photo-based meal scanning, goal-adapted meal suggestions and a family mode.