Inspiration
Medical professionals often struggle with navigating complex, fragmented protocols during critical moments. Traditional search is slow and doesn't always yield actionable results. We wanted to build a tool that not only finds the right protocol but transforms it into a dynamic, context-aware checklist.
What it does
ProCheck is an intelligent platform that uses hybrid search to help healthcare workers find medical protocols instantly. It goes beyond simple search by generating interactive checklists, providing AI-driven query enhancements for better accuracy, and maintaining conversation context for follow-up questions. Every answer is backed by smart citations and confidence scores.
How we built it
The frontend is built with React and Vite, utilizing TailwindCSS and shadcn/ui for a premium, responsive interface. The backend is a FastAPI server that coordinates between Elasticsearch (for hybrid BM25 + Vector search), Google Gemini AI (for embeddings and response generation), and Firebase (for secure authentication and data storage).
Challenges we ran into
One of the biggest hurdles was implementing the Reciprocal Rank Fusion (RRF) for hybrid search to balance keyword relevance and semantic meaning. We also focused heavily on performance, implementing virtualized scrolling and optimistic updates to ensure a smooth experience even with extensive medical data and long chat histories.
Accomplishments that we're proud of
We successfully integrated a robust hybrid search system that significantly improves retrieval accuracy over standard methods. We're also proud of the protocol generation system, which accurately extracts and formats actionable steps from unstructured medical documents.
What we learned
We gained deep insights into the nuances of medical data indexing and the importance of citations in building trust with AI. Tuning Elasticsearch for specific medical terminologies using Gemini-enhanced queries was also a significant learning curve.
What's next for ProCheck
We plan to add multi-language support to make the platform accessible globally. Future updates will also include a mobile app for on-the-go access, direct integration with EHR systems, and voice commands to allow for hands-free operation in clinical environments.
Log in or sign up for Devpost to join the conversation.