QResp

Project overview

This is a project for the Qresp challenge, a web application accessible via a QR code design to assist MPID patients in case of a medical emergency.

It offers a simple and user-friendly interface, that allows both patients and medical care workers to easily provide patient data, medical history and currents signs and symptoms.

By using an advanced knowledge-based system and the given algorithm, the system will respond with crystal clear instruction on what to do in such cases, even for non-specialized medical care workers.

Autonomous update system

Additionally, we created a knowledge database that keeps itself updated on the most recent and relevant research papers related to MPID. It allows for users to query information about it and the system will respond with new and updated information.

Techstack

For this project we used a very solid foundation, based primarily on Python. We built advanced algorithm that allowed the system to easily process incoming data and retrieve relevant extracts.

In order to archive all this, we used OpenAI embeddings in order to be able to compute similarities between different pieces of documentation, which allowed a much faster and most accurate search. Results are saved into QDrant (a vector database), which aid us in approximate neighbor search using hnsw (Hierarchical navigable small world).

A Large Language Model was used to process extract information and create a user-friendly response.