AI Health Tracker

AI-powered nutrition assistant that transforms meal tracking into personalized health insights, summaries, and smarter daily habits.

Comment

Inspiration

Most nutrition apps fail for one simple reason: tracking food manually is frustrating.

People want to eat healthier, but they do not want to spend time entering every calorie, ingredient, or meal into spreadsheets and forms. Logging meals becomes repetitive, slow, and easy to abandon.

We wanted to build a product that removes friction from healthy living. Instead of forcing users to manually track everything, we use AI to make nutrition guidance faster, simpler, and more useful.

Our goal was not just calorie counting , it was helping people build better habits over time.

⸻

What it does

AI Health Tracker helps users understand their nutrition and behavior patterns through AI-powered meal logging and summaries.

Core Features

  • Meal Tracking by Text Users can describe what they ate in natural language, and the AI estimates calories, macros, and nutritional information.
  • Meal Tracking by Photo (prototype direction) Users can upload a meal image for automatic food understanding and nutrition estimation.
  • Automatic Data Storage Structured meal records are saved into the database for long-term tracking.
  • Daily Summary Users receive a quick overview of calories, nutrition balance, and daily progress.
  • Weekly Habit Analysis The system reviews patterns over time, helping users identify consistency, overeating trends, or missed goals.
  • AI Chat Assistant Users can ask questions and receive recommendations related to nutrition and healthier choices.

Our vision is to move beyond simple logging and toward intelligent habit coaching.

⸻

How we built it

We designed the project as a full-stack multi-service architecture.

Frontend

  • React
  • TypeScript
  • Tailwind CSS

Backend

  • Java Spring Boot
  • REST API
  • Business logic
  • Database integration

AI Service

A dedicated Python FastAPI microservice powers AI functionality:

  • meal understanding
  • nutrition estimation
  • daily summaries
  • weekly summaries
  • chat recommendations

Database

  • PostgreSQL

This separation allows the product to scale cleanly and lets the AI layer evolve independently from the main backend.

⸻

Why we choose Gemma 4

We selected Gemma 4 because it matched our product goals beyond simple text generation.

Strong reasoning potential

We wanted a model that could help analyze behavior patterns over time , not only identify food items.

Efficient and scalable

Consumer health products need cost-efficient inference and practical deployment options.

Privacy-focused future

Health data is sensitive. Open model ecosystems create future opportunities for private, self-hosted, or on-device AI experiences.

Multimodal direction

Gemma aligns well with future text + image workflows for meal understanding.

⸻

Challenges we ran into

Hackathons move fast, and we focused heavily on architecture and backend delivery.

Some features were not fully polished by submission time:

  • Photo upload flow was incomplete
  • Chat functionality needed additional backend refinement
  • Free-tier model APIs introduced slow responses
  • Integrating multiple services under time pressure was challenging

Even with those constraints, we delivered a real working system and production-oriented architecture.

⸻

Accomplishments that we’re proud of

  • Built a full-stack product in hackathon time
  • Created separate AI microservice architecture instead of a single demo prompt
  • Connected frontend, backend, database, and AI systems
  • Designed long-term product direction, not only a short-term demo
  • Delivered habit-analysis features beyond calorie counting

⸻

What we learned

  • How to divide responsibilities across a technical team
  • How to integrate frontend, backend, database, and AI services quickly
  • How challenging real AI latency, APIs, and production constraints can be
  • How to design systems that can continue growing after the hackathon

⸻

What’s next for AI Health Tracker

Product Improvements

  • Faster responses through optimized/local inference
  • Fully completed meal photo analysis
  • Better chat assistant experience
  • Cleaner UI/UX flows

Health Ecosystem Expansion

  • Wearable integrations
  • Personalized coaching plans
  • Goal tracking
  • Smart reminders

Privacy & Deployment

  • Explore self-hosted/open-weight deployment paths
  • Improve security and user data ownership

Mobile Experience

  • Native mobile app for faster photo logging and daily usage

⸻

Built With

Gemma 4, Python, FastAPI, Java, Spring Boot, PostgreSQL, React, TypeScript, Tailwind CSS

⸻

Final Vision

AI Health Tracker is not just a food logger.

It is an early step toward a future where AI helps people understand habits, make better daily decisions, and live healthier lives with less friction.

Built With

Share this project:

Updates