PhD research focused on developing and validating machine learning pipelines for large-scale, real-world physiological and multimodal data. My work spans representation learning, statistical modeling, and end-to-end software development, with an emphasis on robustness and evaluation on noisy datasets.

In my primary project, I developed a diagnostic system for seizure localization using intracranial EEG, combining dynamical network models and spectral entropy without supervised training. The method achieved 97.43% sensitivity and 92.48% precision over 1,000+ hours of patient data and provided objective, data-driven support for clinical decision-making.