Research & Projects
A selection of my research and engineering work.
Physics-Informed Neural Networks (PINNs)
Advisor: Prof. Yaoqing Yang | Dartmouth College
Investigating the core properties of PINNs to better understand their training dynamics and predictive capabilities. My work involves analyzing phase transitions and loss landscapes to improve model convergence, as well as studying the impact of floating-point precision in optimizers like L-BFGS.
GPT-Enhanced PINNs for Solving PDEs
Advisor: Prof. Zhichao PENG | HKUST
Addressing key training challenges in Physics-Informed Neural Networks for transport phenomena by combining meta-learning with a Lagrangian perspective. My research focuses on leveraging the strengths of both GPT-PINN and Lagrangian PINNs to enhance model efficiency, accuracy, and convergence for parametric systems.
Deep Learning for MRI Reconstruction
Advisor: Prof. Hao Chen | HKUST UROP
Assisted in training deep learning models to accelerate medical image analysis. My primary contribution was focused on organizing datasets and using convolutional neural networks for the rapid reconstruction of 3D human models from 2D MRI slices.
HKUST RoboMaster Team
Software Engineer
As a member of the university robotics team, I developed a computer vision system for QR code scanning and target identification using Python and OpenCV. I also programmed low-level motor controls and sensor integrations in C++ for our autonomous robot.