MedWise

Your AI-powered medical companion for understanding treatments, analyzing drug interactions, and making sense of complex research — fast, accurate, and patient-friendly.

Comment

🌟 Inspiration

MedWise was born out of the frustration of trying to make sense of complex medical research and drug information, whether you're a clinician under pressure or a patient overwhelmed with jargon. Each year, over 100,000 people in the U.S. and Canada die from preventable drug interactions. I wanted to build a tool that bridges the gap between trusted research and everyday understanding, empowering users to make informed health decisions with clarity and confidence.

💡 What It Does

MedWise is an AI-powered medical research assistant that enables users to:

  • Analyze potential drug interactions using LLM reasoning
  • Retrieve the latest treatment insights from PubMed-style medical literature via semantic search
  • Understand complex medical terms through simplified, patient-friendly language
  • Cite trustworthy sources with confidence scores linked to original research

It features two modes:

  • Clinical Mode for professionals seeking fast, evidence-based insights
  • Novice Mode ("Virtual Patient Mode") for patients who need clear, digestible explanations

🧱 How We Built It

  • Frontend: Built using React, styled with clean and trustworthy UI components and Lucide icons.
  • Database: Used MongoDB Atlas to store structured PubMed abstracts and FDA drug data.
  • AI:
    • Vertex AI Embeddings API was used to convert medical texts and user queries into vector embeddings.
    • MongoDB Atlas Vector Search powers semantic retrieval of relevant papers and drug data.
    • Gemini API (Generative AI) synthesizes findings, analyzes drug interactions, and provides plain-language explanations.
  • Architecture: Designed with modularity, speed, and clarity in mind — deployable via Google Cloud Functions or Streamlit.

🚧 Challenges We Ran Into

  • Solo sprint by choice: As a solo developer, I intentionally challenged myself to build the full-stack project independently — from AI integration to frontend design. It was intense, but deeply rewarding.
  • Medical complexity: Handling sensitive data while ensuring accuracy and usability posed ethical and technical challenges.
  • Multiservice integration: Orchestrating MongoDB, Google Cloud AI, and a responsive frontend within a hackathon timeframe demanded fast learning and creative workarounds.

🏆 Accomplishments That We’re Proud Of

  • Designed a clean, user-trusting UI that feels professional and intuitive
  • Built a fully working AI retrieval and reasoning pipeline using real datasets
  • Created dual-use modes for both clinical professionals and novice users — a rare UX consideration in medical tools
  • Demonstrated the feasibility of semantic vector search + LLMs for critical healthcare use cases

📚 What We Learned

  • How to generate and utilize vector embeddings for semantic search with MongoDB Atlas
  • How to structure prompts and workflows with Gemini to ensure trustworthy, explainable outputs
  • How to scope and deliver an impactful MVP solo under tight deadlines
  • That thoughtful UX and AI together can make a massive difference in accessibility and health equity

🔮 What’s Next for MedWise

  • Live PubMed integration: Connect to APIs for real-time literature updates
  • EHR/EMR integration: Allow doctors to analyze interactions based on actual patient profiles
  • Voice assistant mode for accessibility
  • Mobile-first deployment for broader reach
  • Patient-provider collaboration features: Export findings, flag questions for follow-up
Share this project:

Updates