PhD student · Computer Science · Vanderbilt University
(Nulixiati Mangnike)
Machine learning for physics simulation and computer graphics.
I build learning-based methods for fluid simulation, neural representations of simulation data, and generative models for 3D content, advised by David Hyde.
I am a PhD student in computer science at Vanderbilt University (expected May 2027), advised by Professor David Hyde, where I work as a research and teaching assistant. Before Vanderbilt, I earned my bachelor’s degree in computer science at Peking University. In summer 2025 I was a software engineering intern at Amazon Web Services (AWS).
My research combines machine learning with computer graphics and computational physics. Recent work includes a size-aware 3D virtual try-on pipeline built on 3D Gaussian Splatting, implicit neural representations for compressing high-resolution simulation data, physics-informed networks that improve Boussinesq flow models using compressible-flow simulations, diffusion models for 3D shape synthesis, and physics-grounded video generation. I am also interested in parallel computing and 3D computer vision. For more detail, see my resume.
Outside of research I enjoy football (soccer), drawing, cooking, photography, bodybuilding, reading and hiking.
Physics-informed neural networks that learn corrective dynamics from fully compressible simulations to improve lower-fidelity Boussinesq models, alongside implicit monolithic mixed finite element solvers for Boussinesq and compressible flow.
Implicit neural representations that compress high-resolution 2D and 3D simulation fields while preserving fine-scale structure, fed by large-scale synthetic data pipelines built on parallel FEM solvers (FEniCS).
A size-aware, photorealistic garment transfer pipeline that uses 3D Gaussian Splatting for garment reconstruction and rendering.
Diffusion models for 3D shape synthesis across voxel, point-based and implicit representations, and physics-grounded video generation by fine-tuning vision-language models with geometric and physical priors.
Journal of Computational Physics
Platform for Advanced Scientific Computing (PASC), 2024
Vanderbilt University
Machine learning for physics simulation and computer graphics with Prof. David Hyde: physics-informed networks for Boussinesq flow, implicit neural representations for simulation data, 3D Gaussian Splatting for virtual try-on, and generative models for 3D shapes and video.
Amazon Web Services (AWS)
Built a user-facing analytics console for real-time visualization of deployment and usage data, backed by a secure full-stack pipeline spanning ingestion, APIs and console UI.
University of California, Davis
Implemented differentiable elasticity simulations with FEM and Position-Based Dynamics for learning-based control and optimization; explored the Material Point Method for fracture simulation.
Institute of Software, Chinese Academy of Sciences
Proposed a semi-analytical surface tension model and integrated it into a GPU-accelerated, CUDA-based SPH pipeline for real-time fluid simulation; contributed to the PeriDyno simulation engine.
Peking University
Implemented mesh and topology optimization for hyper-elasticity simulation; contributed to the PhysIKA physics-based simulation engine.