research/

my research sits between machine learning and real world data, where i build methods that turn complex, high dimensional signals into reliable predictions and decisions. i care about problems like early signal detection, robustness under noise and missing data, and making model outputs usable, not just impressive. some of my work is grounded in biomed settings, but the goal is broader: bridge messy data to actionable insights.

my approach leverages modern ml architectures, most recently attention based models and graph neural networks, to learn structure in heterogeneous datasets. i build multimodal methods that combine complementary signals into unified representations, then validate them through careful evaluation and deployment minded pipelines. the bottom line stays the same: learn from diverse data, surface signal early, and translate it into outputs that can be trusted.