I’m a final year PhD student in Machine Learning at University of Cambridge and Max Planck Institute for Intelligent Systems under the Cambridge-Tübingen PhD Fellowship, advised by Prof. José Miguel Hernández-Lobato and Prof. Bernhard Schölkopf.

I’m keen on core machine learning research and its scientific applications. My current research focuses on deep learning and probabilistic methods for generative modeling and sampling (using diffusion and flow models in particular), with applications to image, video and molecular conformation generation. I’m particularly interested in developing deep generative models for photo-realistic visual generation and designing neural samplers for efficient protein dynamics emulation.

Previously, I worked on robust and data-efficient probabilistic models with meta-learning and multi-task learning for molecular property prediction. In addition, I have also been investigating the training dynamics of neural networks, including analyzing the characteristic activations of ReLU neural networks, developing optimal client sampling schemes for federated optimization of neural networks, and examining the behaviors in joint training of deep ensembles.