🩺 ##MedaskAI – Your AI-Powered Health Companion

🚀 Inspiration The healthcare world is vast and often hard to navigate—especially for individuals trying to understand their prescriptions, search for natural remedies, or track personal symptoms. The lack of accessible and AI-powered personal health tools inspired us to build MedaskAI — a user-friendly, AI-enhanced health platform that bridges traditional knowledge and modern tech.

🧠 What We Learned MongoDB Atlas Search is not just a database tool — it's a powerful engine for relevance-based search and vector queries when combined with AI.

Google Cloud’s Gemini AI and Vision API can rapidly extract and interpret real-world medical documents like prescriptions, offering real-time personalized insights.

Integrating Supabase for authentication and journaling gave us a clean and scalable user management system.

Full-stack deployment on Google Cloud Run taught us how to ship scalable apps efficiently with isolated services.

🛠️ How We Built It Frontend:

Built with React 18, Vite, TailwindCSS, and shadcn/ui

Features include: Natural Remedy Search, AI Chat, Prescription Upload, and Health Journal

Backend:

Node.js + Express API

Google Cloud Vision for OCR on prescription images

Google Gemini AI for generating patient-friendly summaries

MongoDB Atlas for storing and indexing 500+ natural remedies

Supabase for secure user authentication and health journal storage

Deployment:

Both frontend and backend are containerized and deployed on Google Cloud Run

🧗 Challenges We Faced Fine-tuning Gemini prompts to generate medically accurate and readable summaries.

Uploading and configuring large datasets (natural remedies) with MongoDB Atlas Search Indexes.

Ensuring AI-generated data matches our UI expectations and keeping latency low on GCP.

Managing real-time OCR → AI summary flow for images within size and performance constraints.

MedaskAI is not just a hackathon project — it's our vision for how AI and open data can empower personal healthcare.

Built With

Share this project:

Updates