Research Statement

My research aims to develop language models whose latent spaces encode information analogous to Plato's Theory of Forms. I envision creating neural architectures where the latent representation of a concept contains its essential nature, similar to how Platonic Forms embody the essence of objects. In this framework, different projections of the same latent embedding would generate various manifestations related to the concept—whether descriptions, actions, or relationships. This approach could potentially enhance models' generalization capabilities, enabling them to reason about concepts at a more abstract level while producing diverse but coherent outputs from a unified representation.

During my master's at Drexel University, I collaborated with Korkut Lab at MD Anderson Cancer Center on cutting-edge computational oncology research. I developed machine learning frameworks for predicting responses to targeted cancer therapies by integrating multi-omics data to model therapy-induced proteome dynamics. This interdisciplinary work enhanced my expertise in processing complex multi-omics datasets, designing neural architectures for healthcare applications, and bridging computational methods with clinical needs. The collaboration research thought me novel methodologies for cancer treatment prediction and provided invaluable experience in translational research that directly impacts patient care.