LI Yihai

Research · Computing · Engineering

LI Yihai.

High-Performance Computing & Scientific ML

Hi! I'm focusing on parallel & distributed computing, GPU acceleration, and physics-informed PDE solvers. Previously a TA for Engineering Mathematics and Discrete Mathematics at Trinity College Dublin.

Download CVContact

Open to PhD & Roles

Experience

Where I've built and shipped.

Research Experience

Researcher @ Supervised by Prof. Pavlos Protopapas

Jan. 2021 — Mar. 2021 • Remote Project
  • Acquired algorithms of machine learning specifically in CNNs accelerated by CUDA.
  • Built Proxy model for CNNs’ evaluation based on Bootstrap algorithm in Python using Tensorflow 2.x.
  • Utilized of Grad-CAM Map, Saliency Map to check the internal learning process in Python and Matlab.
  • Drafted and typeset a comprehensive final report in Latex, including detailed performance graphs.

Undergraduate Researcher @ Shanxi University

Nov. 2017 — Nov. 2018 • Taiyuan, China
  • Authored 1-Good-Neighbor Diagnosability of Unidirectional K-ary N-Cubes under the PMC Model in Computer Engineering and Applications.
  • Employed combinatorial proof techniques using mathematical induction and constructive labeling to derive the diagnosability bound for K-ary N-Cubes.
  • Coordinated a 5-member team over a year, producing one journal submission on schedule.
  • Cultivated advanced skills in formal proof techniques, MATLAB coding, and team project management.
  • Formally derived 1-good-neighbor diagnosability bound k(n−1)k(n-1), and kn−1kn-1.

Internship Experience

Software Engineer Intern @ Oracle

Winter 2018 • Peking, China
  • Assembled X-shell client to control Ubuntu server remotely.
  • Configured and managed encrypted Hadoop clusters on Ubuntu servers via X-shell for data processing.
  • Setup Hadoop files system on Apache to optimize cluster performance and security.

Leadership & Teaching

Teaching Assistant @ TCD

2024 — 2024 • Dublin, Ireland
  • Led lab sessions for C++ and parallel programming; designed reproducible HPC exercises.

Teaching Assistant @ Oracle

Winter 2018 • Peking, China
  • Supported instructors in course delivery and provided guidance to students during big data training sessions.

Selected Projects

A few favorites and proofs of work.

Discrete Fourier Transform Based Convolutional Neural Networks for Visual Recognition

Implemented convolutional neural networks (CNNs) in PyTorch and accelerated training using CUDA-enabled GPUs, detailing performance difference by comparing with custom edge-oriented filters CNNs.

PythonPytorchCUDACNNsHistogram of GradientsFast Fourier Transform
Project overview

Undergraduate Mathematical Modeling Group Competition

Implemented convolutional neural networks (CNNs) in PyTorch and accelerated training using CUDA-enabled GPUs, detailing performance difference by comparing with custom edge-oriented filters CNNs.

RLISp-minerPytorchDimensional-Analysis
Project overview

Meta-Programming and Hybrid Parallel Strategies for Solving PDEs: An FDM and PINN Comparison

Multigrid-based Navier–Stokes solver with halo exchange optimization; strong-scaling to 256 ranks.

C/C++MatlabMPIOpenMPLibtorchCUDAFinite Difference MultigridPINNs
Source code

Dublin Bike-usage Assessment of Pandemic by Means of Deep Learning

Developed and trained LSTM models to capture spatio-temporal patterns on Dublin city bike usage data in the last decade.

PythonMatlabTensorflow 2.xCUDARNNsLSTMs
Project overview
View all projects

Get in touch

I read every email.

Let's work on something meaningful.

For research, opportunities, or a conversation about computing.

Email me