Intro
Welcome to my website! I am Kosmas, an integrative machine learning researcher at OIST. I currently lead interdisciplinary research projects in the Biological Physics Theory Unit, developing AI methods that bridge neuroscience, biodiversity conservation, and behavioral analysis. My work combines machine vision and bioacoustics to solve real-world problems; from tracking fine-grained zebrafish social interactions to monitoring Okinawa's endemic amphibian species. I specialize in deep learning, evolutionary algorithms, and the development of novel computational tools that enable scientific discovery at scale.
Research Highlights
How do social hierarchies emerge from physical interactions? We address this through high-resolution 3D tracking of zebrafish aggressive behavior, capturing 173 experiments across health and disease states. Our dataset reveals fine-grained dynamics of dominance establishment, from initial confrontation to stable hierarchies, providing a foundation for understanding the physical principles underlying social behavior.
Deep learning enables scalable biodiversity monitoring through automated acoustic species identification. Our FrogAI system analyzes frog mating calls to distinguish Okinawa's endemic and invasive amphibian species, transforming hours of field recordings into real-time conservation data. This bioacoustic approach provides a non-invasive foundation for continuous ecosystem health monitoring and early detection of environmental change.
Published in Nature Scientific Data
Dataset available in Zenodo
