MediGuard AI

Your On-Device OTC Safety Companion“Know before you swallow.”

Comment

Inspiration

Every day, millions of people purchase over-the-counter (OTC) medicines without consulting a healthcare professional. Many don’t realize how certain ingredients can interact with chronic conditions or other drugs. As a computer-science student passionate about health tech, I wanted to make drug safety as simple as checking the weather. With the new Chrome Built-in AI and Gemini Nano, I saw the opportunity to create something that runs entirely on-device, protecting user privacy while offering personalized, intelligent health guidance.

What it does

MediGuard AI is a lightweight Chrome extension that helps users instantly check the safety of any OTC medicine by combining:

  • ⚙️ Gemini Nano (local): Analyzes drug names directly on your device.
  • 🌐 Hybrid Gemini Writer API fallback: adds context from FDA datasets when an internet connection is available.
  • 📸 Image detection (optional): Identifies pills and packages from uploaded photos.
  • 💬 Risk Index: Computes a personalized safety score based on age, gender, and health conditions.
  • 🌍 Multilingual support: Summaries and safety instructions can be simplified or translated with a click.

Everything happens privately inside the browser; no data ever leaves your device unless the hybrid mode is enabled.

How we built it

  • Frontend: HTML / CSS / JavaScript
  • AI Logic: Chrome Built-in AI (Gemini Nano + Writer API + Translator API + Prompt API + Summarize API)
  • APIs: FDA Open Drug Data API, Gemini Writer API (fallback)
  • UI Framework: Tailwind-inspired custom CSS (lightweight, responsive)
  • Deployment: Packaged as a Chrome Manifest V3 extension with offline reasoning.

The extension detects the environment. If Gemini Nano is available, it performs fully local reasoning; otherwise, it calls the Gemini Writer API for hybrid cloud-assisted responses.

Challenges we ran into

  • UI performance: Ensuring a polished, consistent popup design that fits Chrome’s constrained extension layout.
  • Model fallback: Designing a seamless switch between local and cloud Gemini models without breaking user flow.
  • Regulatory data parsing: FDA datasets are complex and unstructured; normalizing them for instant lookup was a challenge.
  • Privacy guarantees: Keeping all health inputs local while maintaining accurate AI reasoning required custom data handling.

Accomplishments that we're proud of

  • ✅ Built a fully local AI safety checker powered by Gemini Nano.
  • ✅ Designed a beautiful, production-ready interface that runs smoothly on Chrome Extensions.
  • ✅ Integrated hybrid AI fallback without latency spikes.
  • ✅ Delivered an end-to-end privacy-first medical-assistant experience using only browser technologies.
  • ✅ Completed functional inference, reasoning, translation, and risk visualization within a 360 px popup.

What we learned

  • How to integrate on-device LLMs (Gemini Nano) within Chrome’s new AI APIs.
  • How prompt structure and temperature tuning affect medical reasoning accuracy.
  • How to design lightweight, accessible interfaces for small screen contexts.
  • How to maintain user trust through local reasoning and transparent UX.

What's next for MediGuard AI

  • 🚀 Voice Mode: enable speech input and spoken responses.
  • 📱 Mobile Chrome port: make it usable on Android Chrome with Gemini Nano Edge.
  • 🧩 Smart Reminders: notify users when two OTC medicines conflict.
  • 🧠 Federated health profiles: optional encrypted cloud sync for multi-device use.

Ultimately, I want MediGuard AI to become the world’s most trusted personal medicine safety companion, accessible to anyone, anywhere, without requiring them to share their data.

Built With

  • api
  • chrome-ai-origin-trail
  • chrome-built-in-ai
  • css
  • fda-drug-data-api
  • gemini-nano
  • html
  • javascript
  • manifest
  • prompt
  • summarizer-api
  • translator-api
  • writer-api
Share this project:

Updates