Portfolio
This page includes both course-related and research projects, all of which are open source and subject to ongoing updates. For details of research projects, please refer to the Publications page.
Computer Graphics: Physics-Based Simulation with JAX [Code 
]
This repository provides a suite of physics-based simulation tools using Google’s deep learning framework JAX, featuring both mass-spring methods for cloth simulation and Material Point Method (MPM) for fluid dynamics. By leveraging JAX’s differentiable programming and JIT compilation, these simulations achieve efficiency and flexibility for seamless integration with deep learning frameworks. More examples including ray tracing rendering and inverse estimation will be updated in the future.
Scientific Machine Learning: Neural-Integrated Meshfree (NIM) method [Code ]
This repository provides a differentiable programming neural integrated meshfree (A.k.a. Differentiable Meshfree) solver, designed for both forward and inverse modeling of inelastic materials. This repository supports the accompanying paper with both data and code.
LLM Pretrain and Fine Tuning: AI generated text detection [Code ]
This repository contains code of a course project for fine-tuning a DeBERTa-Xlarge model to detect AI-generated text. The model is trained using all active layers and 5 cross-validation approach for performance evaluation.
