My research goal is to solve fundamental challenges in modern AI—systematic generalization and data-efficient continual learning—by engineering a novel cognitive architecture inspired by first principles of brain function.
I aim to build autonomous agents that robustly understand and reason about the world by integrating three core mechanisms:
(1) disentangled, object-centric perception of the environment
(2) hierarchical world model formation via memory consolidation
(3) adaptive meta-control for reasoning and planning
My unique strength lies in bridging a decade of research in cognitive neuroscience with hands-on engineering skills in deep learning. This background provides a direct path to implement these neuro-inspired principles not as metaphors, but as concrete computational systems. I believe this approach can contribute significantly to developing advanced machine perception and building the foundations for true embodied cognition.