Inspiration
My inspiration for starting Chiron came from becoming interested in mine and my family's health. The way we live our life today and how we take responsibility for it can determine the many years to come. I see how many people wished they knew more bout their own medical situation and how they could it make it better to live a longer, healthier, happier life. Chiron is the first tool that truly allows people to fully understand their health and know how to act.
What it does
Chiron takes in any medical test result and creates a simple summary of the medical record that includes text and generated sliders and graphs, which are simple to understand for people of any background. Based on the user's medical history (self-inputted) and the results they sent in, we are able to recommend the nearest clinic that will help them with their specific medical issue and fits their insurance plan as well as location. The interface is a chat, where the uploaded medical record launches a conversation with a large language model that is specifically trained to talk about health and medicine. Our product allows people to quickly see what areas of their health they should focus on, and come immediately to the right expert to get the medical issue fixed. It also allows people to monitor their health long-term and compare what changes have happened between the results they have sent in at different time periods.
How we built it
We are using OpenAI's API for our MVP and hosting the interface on gradio and hugging face. We used React for making the preliminary website and have enabled google analytics to to give us feedback on our MVP usage.
Challenges we ran into
Some of the bigger challenges were parsing medical records in different languages and getting the most important details to show up in the chatbot's answer first. We also did not have experience with React, gradio or google analytics before, but learnt their fundamentals and built our website. Another challenge was debugging the system, and trying to break it whilst testing it in order to find its weak points.
Accomplishments that we're proud of
We completed this in a very short time frame of about two days, and are excited the potential product we can achieve in just a few months if we continue at our pace. The LLM we have is efficient, simple and on-point in its messages and is already being tested by a large number of people from our networks.
What we learned
Not to reinvent the wheel. Take what already exists, piece it together and add what is missing yourself. Focus on the priorities of the project - the fundamentals of what makes your idea unique - however, keep note of ideas for your roadmap. Get as much feedback on your idea as possible from different people - even strangers. Let them talk about their opinion and just listen, because they will tell you exactly what they need from your product.
What's next for Chiron
Furthering developing our MVP. Making a bug free and user friendly interface so that large-scale testing can begin. Adding more AI aspects, such as comparing and integrating different types of medical records. Obtaining access to electronic medical records for training a personalised medical AI model.
Log in or sign up for Devpost to join the conversation.