DiagnoSys

Logo

Inspiration

Throughout the pandemic, we saw congestion in Canada’s healthcare system even more than previously. With our free healthcare system, we often anticipate and face long wait times in our hospitals and health facilities. Having faced this ourselves, we wanted to create a solution to this issue not only in Canada but worldwide, that could be implemented and used in real life. Because of this, we created DiagnoSys, a one-stop shop for all these problems.

What it does

DiagnoSys is a program and website that allows users to input their symptoms in order to identify whether going to a doctor is an urgent need and what kind of illness the user may have. It takes in user input, in the form of symptoms and their severities, and then analyses it with the help of a pre-trained version of Cohere (LLM). Once the information is fully examined, a response consisting of possible illnesses and helpful suggestions is sent to the user. In the end, DiagnoSys is just a prototype and an idea. It proves that a large-scale project that uses AI to help the health care systems all across the world.

How we built it

As four first-time beginner high school hackers who had just met at Hack the North, we weren’t sure where to start. After a day of brainstorming, we decided to create DiagnoSys with Python and Taipy. Hearing about Taipy and how it worked inspired us to experiment with it as we were all looking to use Hack the North as an opportunity to try new things and step out of our comfort zones. We quickly started working with the Taipy GUI to get an idea of how it would work with the project. Once we had the basics down, something piqued our interest when doing research on how we could elevate our project: Cohere. Realizing we had tokens at our disposal, we quickly took advantage of what we had. After many hours of work and hundreds of errors later, we reached a point where we could combine the front and back-end code ultimately, creating our final product.

Challenges we ran into

We went through many different challenges when tackling our project. When it came to the idea, we found it difficult to settle on a single topic. We thought about working on something that could work with combating deforestation however we settled on Medicare. Python is a language that none of us are too familiar with, however, we thought it would be a good idea, though daunting, to try something new. This ended up being challenging as we were also working with a Python library, Taipy, that none of us had heard of before the hackathon. To add to that, Taipy, being new, was quite challenging to understand as there is not much to go off of on the internet. When running into problems with styling and their stylekit feature, we simply had to go off of lots of experimental testing. After many articles, documents, and sleepless hours, it ended up working out.

Accomplishments that we're proud of

DiagnoSys was not only about achieving our project goals but also about personal and collective growth. We learned Python and Taipy, explored Cohere's API, and gained valuable experience in working with AI technologies.

What we learned

Throughout our fun yet challenging journey of completing this project, we learned many useful skills. The website was made entirely with the use of Python, which allowed (some of) us to actually learn an entirely new programming language and improve and expand on our previous knowledge of its functions and features. We learned how to use Cohere API and communicate with AI in a project, something that none of us have worked with in the past. We learned how to work with a new library, Taipy, which could be helpful in future projects. Finally, this entire experience taught us that it is important to stay persistent because, despite the problems we faced throughout the event, we still came out with a final product.

What's next for DiagnoSys

Once again, DiagnoSys proves that an online AI-backed medicare system is definitely achievable. In the future, we would like to include a medical database where the software is able to access the historical database of each of its users for further accuracy. We would also like the scale the project, taking advantage of more power language models and making use of larger databases. This would allow machine learning to improve our system and further heighten the abilities of our software.