PhysioAI

AI powered physical therapy education platform. The platform is intended to provide contextualized information around injuries for people who don't have access to medical resources.

Comment

Inspiration

Personal Experience, members of our team struggled with access to physical therapy as they had untreated injuries. We realized that this problem is common for not just college students but those across the world with limited access to healthcare.

What it does

Provide contextualized remedy recommendations regarding physical injuries. These may include strains, sprains, or any general pain people may be facing. Based on a set of questions asked by our model, we are able to inform people about what they can do to remedy the pain.

How we built it

Using React, NodeJS, and GPT

Challenges we ran into

When sending the conversation information to the model, each API call we made requires the entire chat history. This meant that each follow up question the model asked the user, the entire chat history needed to be included within our API call. This drastically increases the token count on each API call. To circumvent this issue, we limited each conversation to a total of 8 messages (4 conversation turns).

Accomplishments that we're proud of

How we tuned the model to provide specific information around user input, learning how to abstract our API key, and learning how to use MaterialUI.

What we learned

Full stack development, prompt engineering, and how to collect the right information needed for suggestions.

What's next for PhysioAI

Implementing live AI coaching for form to ensure users are doing exercises the right way, implementing a calendar so people can stay consistent with their exercises, and a recommendation tab that can suggest adjacent stretches/exercises related to their chat history.

Built With

Share this project:

Updates