Recent Work

NBANeural.net is a web interface built in React to view final score projections and "Against the Spread" predictions for NBA games from my machine learning model.

The model itself is a deep learning neural network built in Python using the PyTorch framework. It draws from advanced data metrics that are calculated manually from every play in every single NBA game.

Over multiple seasons of backtesting, it reached or surpassed a 55% winning percentage against the spread.

As a personal project to develop my first full-stack application, I decided to work with the Spotify API. I noticed that it had functionality for generating playlists based on multiple songs, artists, or genres, yet there was no way to do this within the Spotify app.

I designed an interface to select up to five songs to base a playlist on, tweak custom "audio features" that Spotify supports, and then generate a playlist that can be automatically added to your Spotify library.

At some point during college, I stopped hosting this project. I would love to revive it with improved UX, but unfortunately Spotify completely nerfed the ability for independent developers to create new apps using their API. To set up a new app using the Spotify Web API and have it be used by more than 5 people, you need to be an organization with 250k+ monthly active users. Very sad!

The code is still available on GitHub if you really love reading PHP.