What If Your Health App Actually Helped?

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass

What We Built
HealthSync is an AI-powered personal health companion designed to help people step away from screens, go outdoors, stay active, and build healthier lifestyles. It encourages users to walk, run, exercise, and engage with their surroundings and communities instead of spending all their time indoors on laptops and phones. With features like activity tracking, real-time exercise guidance, personalized health assessments, nutrition analysis, and AI-powered coaching, HealthSync helps users understand their health and make better everyday choices. By combining smart technology with real-world physical activity, HealthSync aims to make fitness more accessible, encourage people to touch grass, and create a healthier balance between technology and life.

Demo
https://youtu.be/Rc0mtDuJRjY

Code
https://github.com/Wahaj-udn/HealthSync-Team-IQOOL

How We Built It
Open-weight AI model: Qwen3-4B, a locally runnable language model used for offline health coaching and personalized nutrition analysis.
Local inference: Integrated llama.cpp through a JNI bridge (C++ and Kotlin) to run the quantized Qwen3-4B model directly on Android devices.
AI frameworks: ONNX Runtime for on-device health risk prediction, MediaPipe for real-time exercise pose detection, and Google ML Kit for nutrition-label OCR.
How the project is built around AI: HealthSync combines these AI components to assess health risks, count exercise repetitions, analyze food labels, and provide personalized guidance while keeping most processing on-device.
Privacy-first approach: Local AI enables offline coaching and nutrition analysis without sending that data to external servers. Gemini API is an optional cloud-based conversational coach.
Real-world impact: AI encourages users to move beyond screens by supporting walking, running, exercise, and healthier daily habits.

Why Does Open Innovation Matter?
Privacy and control: Open-weight models like Qwen3-4B let us run AI directly on users’ devices, reducing dependence on cloud services and external APIs.
Offline accessibility: Local inference enables health coaching and nutrition analysis without requiring a constant internet connection.
Customization: Open-source tools like llama.cpp, ONNX Runtime, and MediaPipe let us customize and integrate AI for health assessments, exercise tracking, and personalized guidance.
Cost efficiency: Local processing reduces recurring API costs and makes the app more sustainable to scale.
Our vision: Open innovation helps us build accessible, privacy-conscious health technology that encourages people to step away from screens, get outdoors, exercise, and build healthier lifestyles.

A huge shoutout to my teammate Wahaj-https://dev.to/wahaj-udn

And Thank your DEV Team for considering our submission