Welcome to Yangshuai Wang's homepage

Bio:

July 2024 - Present: Peng Tsu Ann Asssitant Professor at National University of Singapore, mentored by Prof. Weizhu Bao.

Dec. 2021 - July 2024: Postdoc at University of British Columbia, supervised by Prof. Christoph Ortner.

Sep. 2016 - Dec. 2021: Ph.D at Shanghai Jiao Tong University, supervised by Prof. Lei Zhang.

Sep. 2012 - Jun. 2016: B.S at Sichuan University, supervised by Prof. Hao Wang.

Research Interests:

My research is situated at the intersection of Numerical Analysis, Scientific Computing, and Computational Materials Science. I focus on developing mathematically rigorous and computationally efficient methods for molecular modeling and multi-scale simulation. My specific research directions include:

1. AI for Science (AI4Science) & Foundation Models

• LLMs for Scientific Discovery: Investigating the integration of Large Language Models (LLMs) and specialized foundation models to accelerate scientific workflows, automate property prediction, and facilitate "inverse design" in materials science.

• Scientific & Quantum Machine Learning: Developing data-driven solvers and Quantum Machine Learning (QML) techniques for solving high-dimensional PDEs. My work emphasizes the development of robust numerical algorithms with a focus on Uncertainty Quantification (UQ) and the theoretical analysis of convergence and generalization.

2. Machine-Learned Interatomic Potentials (MLIPs)

• Numerical Analysis of MLIPs: Advancing the mathematical foundation of interatomic potentials by investigating approximation errors, stability, and the reliability of force-field predictions.

• Multi-Fidelity & Coarse-Graining: Developing systematic approaches for fine-tuning pre-trained models and constructing coarse-grained (CG) force fields for soft-matter systems.

• Applications in Materials: Applying these potentials to study complex phenomena in energy storage and conversion materials, ensuring a seamless integration of physics-based priors with data-driven flexibility.

3. Multi-scale Modeling of Crystalline Defects

• Coupling Methods: Engineering and analyzing Atomistic-to-Continuum (A/C) and Quantum Mechanics/Molecular Mechanics (QM/MM) coupling schemes.

• Rigorous Error Estimation: Implementing both a priori and a posteriori error analysis to guide the development of adaptive algorithms that balance accuracy and computational cost.

• Boundary Conditions: Formulating sophisticated boundary conditions to accurately capture the long-range strain fields associated with crystalline defects.

4. Microstructure Evolution & Computational Materials Design

• Data-Driven Microstructure Modeling: Utilizing machine learning to decipher the kinetics of microstructure evolution in complex systems, including optical and magnetic materials.

• Multi-scale Property Prediction: Intersection of multiscale behavior and surrogate modeling to predict macroscopic material performance from mesoscopic morphological data, facilitating the accelerated design of high-performance materials.

Contact:

Email: yswang@nus.edu.sg

Address: S17-05-16, 10 Lower Kent Ridge Road, National University of Singapore, Singapore 119076.