Projects

Smaller projects

Course and side projects where I learned to work with data.

Web app · 2024 – 2026

Hitting the Vibe

My first data project · began as an NYU course project (team of four), later built as a working app

An app that picks songs from guests' Spotify playlists and builds one shared party playlist of danceable, energetic tracks that fit together.

How it works. Each guest's songs get a party score (danceability + energy − acousticness). The top tracks are compared by cosine similarity over six audio features, and the 25 that fit the group best form the playlist.
What testing showed. Audio features alone cannot pick out party songs (AUC 0.48–0.60), and standardizing the features first swaps 10 of the 25 songs.
Note. The demo uses simulated guests and a public 2020 Spotify dataset, because Spotify's audio-features API is closed to new apps.

Course project · Summer 2026

Data Mining Methodology II, UCF · team of two

A model with 75.9% accuracy found only 23% of glioma cases. The per-class numbers told a different story from the overall score.

Method. Reduced the pixel features of 3,264 MRI scans in four classes with F-test selection and PCA, then compared nine models, from regularized regression and tree ensembles to a CNN.