Distributed graph pipeline on AWS EMR, turning 2.4M Twitter edges into a low-latency Flask API and React dashboard.

Semi-supervised ensemble of six detectors over temporal features, lifting heavy-truck fault detection by 70% AP.

Diagnosed and repaired an EEG-conditioned diffusion pipeline, lifting 50-way generative accuracy from 0.02 to 0.94.

Multiple linear regression on NHANES survey data, quantifying the demographic drivers of systolic blood pressure.

Post-stratified logistic regression over survey and census data, forecasting vote shares for the 2025 Canadian election.