DubCare

Inspiration

We all had issues of either our parents, our ourselves taking unneeded time and effort to be able to contact hospitals to set up appointments. We recognize the time of doctors and nurses are incredibly valuable, and we want to not take away from that, but improve their workflow by using AI to automate the tasks of making an appointment or receiving information about prescriptions, etc.

What it does

The demo uses ChatGPT and Azure Frameworks to show that a human making an appointment, for example a call, to then be converted into text to be used by ChatGPT. As we are using ChatGPT for demo purposes only, and that future language models such as these will be trained on the hospital datasets, we just wanted to show the capability they have when processing information, which will be much more efficient in terms of benefiting the user. Finally, we produce the output as an audio file, showing that this entire process is fully possible and can greatly benefit the tedious process of having to wait to be connected to a nurse to be able to ask information or schedule an appointment.

How we built it

We built it using Python, in a frontend framework that we actually saw featured on the DubHacks site called StreamLit. We found it very beneficial to use and utilized it to display all our information. For our backend, we utilized the ChatGPT API, as well as Azure Speech Recognition and Speech Recognition from PyPi to be able to accurately process information from not only the user, but the AI itself.

Challenges we ran into

We ran into issues such as audio desync and problems with the AI not being able to efficiently parse the information. We resolved this through experimenting with different frameworks and refining the program so that we could streamline the user experience in interacting with the bot. By doing this, we solved not only the audio issue, but also made further processing of the audio by the bot much more performant.

Accomplishments that we're proud of

We're proud of creating a working proof of concept in 24 hours, which showcases how automating and utilizing AI for these tasks is a very possible solution in our lifetimes.

What we learned

We gained valuable experience about Microsoft Azure and OpenAI API utilization. We also learned to better collaboratively learn and grow as a team to overcome issues in a short timespan.