About Me
AI4Science Research
As a maintainer of the MindScience repository, I lead Huawei's flagship AI4Science platform with 6 specialized suites and 60+ AI models. My research focuses on physics-informed neural networks, scientific computing, and AI-driven solutions for complex scientific problems.
My work spans computational fluid dynamics, spatiotemporal modeling, and physics-encoded machine learning, achieving breakthrough performance in weather forecasting and CFD simulations with 1000x improvements over traditional methods.
AI4Science Platform Maintainer
6 Specialized Suites • 60+ AI Models
LLM Serving Expert
vLLM & vLLM-Ascend Specialist
LLM Serving & Infrastructure
Specialized expertise in large language model serving optimization with deep focus on vLLM and vLLM-Ascend systems. Rich experience with Ascend NPU architecture, optimization techniques, and hardware acceleration. I lead research in high-performance LLM deployment, inference optimization, and distributed serving architectures for production-scale applications.
My work encompasses performance optimization, CUDA acceleration, distributed systems design, and ML infrastructure development, bridging cutting-edge research with practical implementations for enterprise deployment.
Core Technologies
Professional Experience
Research Scientist
Huawei 2012 Lab • Present
1. Maintainer of MindSpore Science open source project, leading AI4Science initiatives for scientific computing and discovery.
2. Active contributor to vLLM & vLLM-Ascend and other LLM serving open source projects.
Ph.D. in Statistics and Operations Research
University of North Carolina at Chapel Hill • 2015-2020
Doctoral research in Statistics and Operations Research, focusing on optimization theory and computational methods. Developed expertise in convergence analysis for decomposition algorithms and large-scale optimization problems.
Bachelor's Degree in Mathematical Statistics
School for the Gifted Young, USTC • 2011-2015
Undergraduate education in Mathematical Statistics at the prestigious School for the Gifted Young at the University of Science and Technology of China, providing accelerated academic training and comprehensive foundation in mathematical theory, statistical methods, and quantitative analysis.
Featured Publications
Showcasing cutting-edge research in AI4Science and ML optimization
Conservation-informed Graph Learning for Spatiotemporal Dynamics
KDD 2025
SlotPi: Physics-informed Object-centric Reasoning Models
KDD 2025
PhyMPGN: Physics-encoded Message Passing Graph Network
ICLR 2025
Featured Projects
Leading initiatives in AI4Science and ML infrastructure
MindScience Platform
Lead maintainer of Huawei's flagship AI4Science platform with 6 specialized suites and 60+ AI models.
LLM Serving Research
Specialized research in large language model serving optimization with vLLM and vLLM-Ascend systems.
MindFlow - CFD Suite
Computational fluid dynamics solver with physics-informed AI and differentiable programming.