Logo + Project Name
Review Section
Upload Section
Description / How it works
More Information Horizontal Bar
Hero Section

Inspiration

A lab report can be full of numbers, ranges, and medical terms while leaving the person reading it unsure what to ask next. We built MediSense to make that first reading less overwhelming and help people have a more informed conversation with a clinician.

What it does

MediSense accepts a lab report or prescription as a PDF or image and turns it into an explanation in everyday language. It highlights measurements that may need attention, keeps prescription details separate from general medicine information, and suggests questions to bring to an appointment. Users can revisit recent scans or save an explanation as a PDF. Illustrative sample reports let anyone explore the experience without uploading a document.

MediSense is an informational aid, not a diagnosis or a replacement for medical advice.

How we built it

We built the interface with React and Vite, and the API with FastAPI. The backend checks each upload’s format, size, and readability before sending it to Nemotron for structured analysis. Pydantic validates the response, and SQLAlchemy saves successful results in SQLite or PostgreSQL. The interface presents the findings in a readable layout with responsive design and reduced-motion support.

Challenges we ran into

The hardest part was making an AI response useful without presenting guesses as facts. Medical documents vary widely in layout and quality, so we designed the analysis to preserve values and directions as written, flag uncertainty, and reject unreadable documents. We also had to handle failed API requests and database saves clearly so users would know when an explanation was not saved.

Accomplishments that we're proud of

We’re proud of the complete journey from upload to a clear, revisitable explanation. MediSense keeps extracted document details distinct from AI-generated context, offers useful sample reports, and gives users practical ways to print, save, and discuss their results.

What we learned

Clarity takes more than simplifying terminology. People need to see which information came from their document, where uncertainty remains, and what questions are worth asking a professional. We also learned how much careful validation and error handling matter when building around AI-generated information.

What's next for MediSense

Next, we want to improve document accuracy across more report formats, add a clear Archive option alongside permanent deletion, and make saved history easier to access across devices. Before inviting people to use MediSense with real medical documents, we also want stronger account access, privacy controls, and production-ready monitoring.