I am an undergraduate researcher at the University of South Florida . My work focuses on data-centric machine learning, particularly data subset and coreset selection. I am interested in understanding which examples are necessary to preserve model behavior, how data should be valued and selected, and how learning can be made more efficient with less data.
Benchmarked knowledge-editing methods on Llama and Qwen models using EasyEdit, analyzing their trade-offs in edit accuracy, generalization, knowledge preservation, and computational efficiency.
Modeled Florida temperature trends and predicted Heat Vulnerability Index (HVI) using geospatial analysis, robust linear regression, and clustering to assess community-level heat risk.
Developed an AI-powered behavioral analysis pipeline that classifies crisis-related Reddit posts by sentiment and risk severity, extracted contextual location signals with NLP, and mapped regional distress patterns through geospatial visualization.
Designed a multi-model equity forecasting pipeline for portfolio optimization, comparing ARIMA, LSTM, and XGBoost across statistical, deep-learning, and gradient-boosting approaches. Performed time-series diagnostics and out-of-sample forecasting to generate predictive signals for downstream portfolio allocation.
I enjoy drawing in my free time. Here are a few selected pieces.