MedTalk

Inspiration

I’ve always felt it's important to understand the medicines we take. Not just what they’re called - but what they do, how to use them properly, and what side effects to watch out for. MedTalk came from that simple idea: help people make sense of their prescriptions without needing a medical degree.

What it does

You upload a prescription image, and MedTalk pulls out the medicine names and details using OCR. Then it gives you clean summaries and lets you ask questions through a smart chat.

How I built it

• Built the frontend with React + Tailwind CSS
• Used Node.js and express for the backend
• Integrated Puter’s OCR API to extract text from images
• Plugged in Google Gemma 3 27b for chat and summaries
• Used Web speech API for speak accessibility

Challenges I ran into

Extracting handwritten and messy printed text was tricky. I tested a few OCR options, but Puter’s API gave better accuracy and fit well with the flow. Also had to keep state stable across chat modes and uploads - it took some debugging to get right.

Accomplishments I'm proud of

• One-click flow from image to chat
• Smart and simple summaries tailored for users
• Branded and responsive UI
• Clean state management with scalable setup

What I learned

• OCR isn’t plug-and-play, it needs testing and tuning
• State logic across dynamic UI needs planning
• Prompt tweaks really shape AI output
• A clean UI helps people trust the product

What's next for MedTalk

• Add voice support and translation
• Build a glossary for medical terms
• Create a clinic version with analytics

Built With

• express.js
• gemma
• node.js
• react
• tailwindcss