Since I began my journey, I've been passionate about connecting mathematics with real-world problems, particularly in life sciences. My interdisciplinary background allows me to address problems across fields. My research focuses on machine‑learning and operator‑theoretic approaches to nonlinear dynamical systems.
My experiences reflect this interdisciplinary focus, supported by a computer science foundation. During undergraduate research, I studied and applied network dynamics theory to model disease spread during COVID‑19. Later, at Johnson & Johnson, I developed AI frameworks to enhance drug discovery and biomedical analysis using deep learning and graph neural networks. My computer science training equips me with powerful tools for computational modelling and data analysis that facilitate further effective collaboration across scientific disciplines. Beyond equations and code, I’m equally interested in the people behind the science—the setbacks they faced and the pivots they made.
Joining the Hong Kong Laureate Forum is an opportunity to engage with laureates and young scientists from many different disciplines. There, I hope to learn from laureates about the choices that shaped their work and exchange experiences with young scientists. My goal is to leave with valuable ideas I can apply to my research, along with mentors and collaborators who will support my growth.

Yudhistira A. Bunjamin - UNSW Sydney

The transfer operator method allows us to convert complex, nonlinear dynamical systems into linear representations that are simpler to analyse. While the spectrum of these operators may contain important insights into system predictability and emergent behaviour, approximating transfer operators from data can be challenging. We address this problem through the lens of general operator and representational learning, in which we approximate the action of transfer operators on infinite-dimensional spaces using tractable finite-dimensional forms. Specifically, we machine learn orthonormal, locally supported basis functions that are tailored to the system dynamics. These learned basis functions and dynamics then serve to compute accurate approximations of the transfer operator's eigenpairs. Our approach offers a promising direction for robust approximation of transfer operators, particularly in high-dimensional systems where traditional numerical methods may become computationally intractable.