Tianshi Xu
Postdoctoral Fellow
Emory University • Department of Mathematics
Email: tianshi ( d o t ) xu ( a t ) emory ( d o t ) edu
Research Interests
My research advances learning-augmented scientific discovery through end-to-end algorithm co-design: structure-aware models, geometry-aware optimization, and hardware-efficient computation integrated in scalable numerical solvers.
Position
- Postdoctoral Fellow, Department of Mathematics, Emory University (Aug 2023-Present, Mentor: Professor Yuanzhe Xi)
Education
- Ph.D. in Computer Science, University of Minnesota, Twin Cities, U.S. (2023, Advisor: Professor Yousef Saad)
Recent News
- [Feb-2026] ACM26 will be hosted at Emory Mathematics. [Website]
- [Jan-2026] Paper accepted by SIAM Journal on Scientific Computing: "Mixed Precision Orthogonalization-Free Projection Methods for Eigenvalue and Singular Value Problems"
- [Jan-2026] Paper accepted by Journal of Open Source Software: "HiGP: A High-Performance Python Package for Gaussian Processes"
- [Dec-2025] Paper accepted by SIAM Journal on Scientific Computing: "Multiscale Neural Networks for Approximating Green's Functions"
- [Nov-2025] New preprint: "Designing Preconditioners for SGD: Local Conditioning, Noise Floors, and Basin Stability"
- [Oct-2025] New preprint: "Preconditioned Truncated Single-Sample Estimators for Scalable Stochastic Optimization"
- [Oct-2025] New preprint: "Neural Approximate Inverse Preconditioners"
- [Apr-2025] New preprint: "Preconditioned Additive Gaussian Processes with Fourier Acceleration"
- [Mar-2025] We are organizing the Atlanta Mathematics of Data Science Bootcamp. [Flyer]