WillowWell

Langflow used to tailor Google Gemini and ran in website.

Inspiration

We understand that finding the right care is both challenging and difficult for many who don't have resources. WillowWell can help reduce that friction and make people feel more comfortable with, and more informed, when receiving care.

What it does

WillowWell has a custom tailored AI that is trained on healthcare data records. It uses this to give a user accurate information that actually applies to them. The user interacts with the AI via a send and receive model.

How we built it

The website is built with Next.js, TypeScript, and tailwind.css. The AI that we integrated uses data from a custom Mongodb database and is tailored through Langflow. The AI interacts with the website via a Langflow API and app router.

Challenges we ran into

The dimensionality of Google Gemini embedding, Mongodb, and our data set was not synchronized. Google Gemini uses 768 dimensions to create relations and classifications, our data only used 5. Because of this discrepancy, we had to fully update our data, creating a JSON file that is over 16000 lines long. Even with over 16000 lines of text, there is not enough data to fully utilize vector based search for our RAG model. There are times the LLM didn't have any context at all to create a response based upon context. For future development, we intend to increase the size of our data for better context analysis.

Accomplishments that we're proud of

We created a personalized RAG model using an open source software (Langflow) that is able to retrieve data from our database and give contextual answers based on input. Moreover, our custom created text boxes allow for pairing deletions and can be scrolled internally.

What we learned

We learned the tools necessary to develop RAG based models that interact with databases. Moreover, we also learned the ability to create chat boxes capable of being scrolled and logged, with deletions of pairs that don't other prompts.

What's next for WillowWell

WillowWell is a startup founded within Carnegie Mellon University's Swartz Center for Entrepreneurship with plans to be also connected with Boston University's innovation incubator. We are currently in the works to complete consumer discovery through meetings with physicians at University of Pittsburgh Medical Center and patients with many different medications. WillowWell seeks to be an industry standard in patient health awareness, and effective communication between the patient and the physician.