I am a researcher in machine learning (ML) and artificial intelligence (AI) at CSIRO’s Data61. My core interests lie in geometric deep learning, information theory, and the intersection between modern ML methods and the foundational field of information geometry.
Over the past decade, I have consistently published in top-tier ML venues including ICML and NeurIPS. I serve as an area chair for ICLR 2024, 2025 and NeurIPS 2025. I am actively engaged with scientific communities in Sydney, Geneva, Grand Paris, and Tsinghua alumni network.
My current research is guided by the following directions:
— To draw upon uncharted mathematical and physics methods that are overlooked by much of the ML world (e.g. those from general relativity and quantum calculus), and to venture into directions that may subtly hint at future ML methods.
— To bridge the gap between ML and information geometry, particularly by adapting fundamental concepts (Fisher information; divergences) that were originally built on simple statistical models to the modern deep learning realm.
— To gain deep perspectives that are counter-intuitive and contrast with common perceptions of ML principles, and generalize these insights into prototypes that are beyond current technologies.
Some of my cornerstone methodologies are
— Build first, even if it seems useless at the moment.
— Many directions pursued with deep commitment can lead to profound and unexpected insights.
