InjuryInspect

🚀 Inspiration

Everyone has experienced using Google to try to diagnose a new source of pain, only to find a slew of unhelpful and occasionally terrifying information. Whether you're a runner with a new pain in your legs, or you just badly rolled your ankle, we always want to know what's going on with our bodies and how we can feel better.

💪 What it does

A user clicks on a point on the human model where they feel pain, and InjuryInspect produces a detailed report helping them diagnose their injury and explore treatment options. InjuryInspect gets its data by searching the Mayo Clinic website for relevant information, and using AI to summarize data into appropriate and digestible points. The user can then chat with our friendly chatbot, to explore their options and learn more about their condition. Of course, the chatbot will not hesitate to instruct the user to contact a medical professional if necessary, as not every condition can be treated from home.

🖋️ How we built it

We used TaiPy, a python web application framework, to build our frontend. Our python-based backend makes use of a custom-built web scraper that searches the Mayo Clinic website for articles, which are then summarized by OpenAI's GPT-3.5 API, thus avoiding the most common problem with AI - outdated and untrue information. InjuryInspect will always be up-to-date and accurate, because it relies on the constantly-updating knowledge of medical professionals, rather than a machine learning model's training data. We also used Cohere's Embeddings API to ensure that InjuryInspect picks the highest-quality unique generations from OpenAI. Our implementation uses the best of two AI technologies as well as our own hand-built systems to create a seamless and thought-out package that is uniquely our own.

🚧 Challenges we ran into

Even though TaiPy says it would "cut development time in half", a simple web application frontend that should have taken a few hours took us almost the entire hackathon. Most things that would be trivial to implement with HTML, CSS, and JS were extremely difficult to pull off with TaiPy. We also had difficulty coming up with a project idea. We bounced back and forth between several ideas, each with their own faults, before finally settling on InjuryInspect, an idea that allowed us to investigate interesting technologies and design some of our own.

🎉 Accomplishments that we're proud of

We overcame the frustrations of TaiPy and managed to finish our project, which ended up more polished than most of our previous hackathon endeavours.

📖 What we learned

Please, for the love of all that is good, come up with some ideas before hacking time starts. We wasted almost 9 hours of our hacking time on idea generation, the stress of losing time making it even more difficult to focus on nailing down the perfect idea. If we had found a project idea a few days, or even a few hours in advance, we would have had much more time to work on our project.