1st Place By Google in Shellhacks Hackathon 2023

Inspiration

Institutions that lack resources and representation might not have the means to access elite sports coaching. Our team holds the view that important knowledge about sports techniques can be curated and presented as a reference for those who cannot afford coaching for their teams. Our aim is to organize information about correct techniques and make it universally available.

What is Athlete AI?

A user will upload a brief video showcasing their form and technique. Our specialized algorithm, utilizing computer vision, will assess their posture and movements. By contrasting this with our established models, the system will produce an analytical breakdown of how their form measures up to the ideal standard for that movement. Based on this data, users will receive recommendations to enhance their form and a visual representation of their wireframe during the action.

1st Place Patient Education Category in UF Designathon 2022

Inspiration

For this Designathon, we were tasked with creating a project that improves patient or healthcare education. When we first started collaborating as a team one of the ideas that came up was Dyslexia, which is one of the most common learning disabilities in children. We were able to find research and a dataset that we thought we use to improve patient education surrounding Dyslexia. Within the United States, about 14.5 to 43.5 million children and adults have dyslexia and we knew we could help raise awareness about it through our project.

What is Dyscovery?

Our website provides information about the condition and predicts if a user is Dyslexic by replicating four distinct word games from a prominent dyslexia medical study. We use indicators from the user playing the games (including clicks, hits, misses, overall accuracy, and time taken) and their demographics to make predictions. Using a Decision Tree machine learning classifier we can predict dyslexia with 85.19% accuracy and sends that result to the user using the FastAPI backend server.

My Contribution

My contrubution was setting up the backend server, training and fitting the machine learning model, and deploying the model onto the server. I utilized the Sci-kit Learn Library to train and fit the model and I set up a Python Runtime Server using FastAPI to deploy our Decision Tree Machine Learning Model (created initially in Jupyter Notebook). I also used Pickle, another Python library, to pre-train our model for fast predictions.

1st Place for Best Overall in UF SwampHacks Hackathon 2023

Inspiration

Having to watch extensive and tedious lectures has consistently been a source of frustration for me since coming to college. Some professors use verbose language, lecture slowly, or display the content in an unengaging manner. I believe that the essential content of these videos could be condensed and distilled into only the most important topics, thereby increasing my productivity. This problem was the inspiration for my team and I's submission to the 2023 SwampHacks Hackathon, the largest student run hackathon at UF with over 300 competing teams.

What is Blue Note?

Blue Note provides an easy way for users to understand video content by summarizing videos and defining the key terms in the video. It allows users to register and log in with ease via Firebase Authentication and input any YouTube video simply by pasting a URL. A generated summary, keywords, and definitions are subsequently displayed to the user and also stored locally for future reference. This allows users to refer back to the information at any time, making it a convenient and efficient tool for watching lectures.

My Contribution

My job was setting up the backend server, API routes, API calls from the frontend, and API calls to the various APIs we implemented. The React frontend is used to accept the URL, video title, and genre from the user. This information is then processed by a Flask backend where the Youtube Transcript API is called to get the transcript of the video. The transcript is then processed by the OpenAI GPT-3 API, and a summary as well as keywords are generated. These keywords are then defined by Wikipedia via the Wikipedia API, and then everything is passed back to the frontend via an API route.