Computer Science at UCL
I analysed about 96,000 delivered orders from the Olist Brazilian e-commerce dataset in PostgreSQL, answering six business questions: revenue growth, regional concentration, late deliveries and reviews, payment methods, repeat customers and spend concentration. I did it to practise writing SQL that…
I built a classifier that predicts whether a protein is an enzyme, and which of the six main enzyme classes it belongs to, from its amino acid sequence alone. I used frozen ESM-2 embeddings with a Logistic Regression and XGBoost ensemble on 39,678 sequences, where 81.5% were non-enzymes. Because…