Extensively explored feature engineering, dimensionality reduction (PCA, t-SNE), and clustering (DBSCAN, kmeans) techniques. Best model created: two consecutive neural networks, an autoencoder and a self-organizing map (SOM) to organize >1000 distinct fonts.
Group capstone project for University of Toronto SCS Machine Learning course - I had the added responsibility as team leader.