CV

Summer Intern (ML Software Engineer)

Meta

Built a feature-importance aggregation API to generate importance scores from historical records, reducing computation time from days to minutes in production models.

• Python, SQL

Summer Intern

International Computer Science Institute

Developed and implemented a C++ algorithm to compute absolute multi-parameter persistent homology, reducing time complexity from $O(n^3)$ to $O(n \log n)$

• C++, Python, Pybind11, CMake, Pytorch

Summer Intern

NSF-Mathematical Sciences Graduate Internship

Developed advanced topological optimization heuristics, integrated as a custom neural network layer in PyTorch, and demonstrated improved loss convergence rates in diverse experiments, including generative AI models and time series data cleaning.

• C++, Python, Pybind11, Pytorch

PhD

University of California, Davis

Applied Mathematics

M.S.

University of Chicago

Computational and Applied Mathematics

B.S.

University of Liverpool

Applied Mathematics

geomp (hidden for paper review)

Parallel geodesic computation and mesh comparison based on geometric energy minimization

• Geometric Energy Minimization by PyTorch
• CPU Parallelization
• Geometry Processing

hyperbolic-shape-morphing (hidden for paper review)

Generate hyperbolic kaleidoscope patterns by shape morphing on the hyperbolic disk

• Energy minimization on the hyperbolic disk

graph-mph0

Efficiently computing 0-dimensional persistent homology on graphs

• C++
• Python
• Pybind11

NLP+RSA

Use Transformer to decrypt encrypted message generated by RSA algorithm

• Transformer
• Built RSA in C++&Python

Torch-TDA

Automatic differentiation of topological features and common topological loss functions

• PyTorch
• Automatic Differentiation

BATS

Computing and visualizing persistent homology

• C++
• Python Bindings

Computing Betti tables and minimal presentations of zero-dimensional persistent homology

Yuan Luo, Dmitriy Morozov and Luis Scoccola

Developed algorithms to compute Betti tables and minimal presentations of zero-dimensional persistent homology

Accelerating iterated persistent homology computations with warm starts

Yuan Luo and Bradley J. Nelson

Developed algorithms to accelerate persistent homology computations based on updating associated matrix factorizations

Topology-Preserving Dimensionality Reduction via Interleaving Optimization

Bradley J. Nelson and Yuan Luo

Developed a topological type-I/II error to identify the correspondence between original and projected datasets