Richard (Hui) Li

University of California, Irvine (UCI), CA
Master of Science in Computer Science
Sept. 2023 - Expected June 2025
GPA 4.0/4.0

Huazhong University of Science and Technology (HUST), Wuhan, China
Bechelor of Engineering in Computer Science and Technology
Sept. 2019 - June 2023
GPA 3.98/4.0 (top 2%)

Machine Learning Research Intern
OPTML Group, Michigan State University
June - Oct. 2022
• Proposed a novel model SMUG that systematically integrates randomized smoothing with the state- of-the-art deep neural network to mitigate the lack of robustness in Magnetic Resonance Imaging (MRI) reconstruction
• Designed a novel unrolling loss that matches the unrolling architecture of SMUG and a “pre-train + fine-tune” training scheme to improve the effectiveness and efficiency of training SMUG
• Conducted extensive experiments using Python and PyTorch and improved peak signal-to-noise ratio (PSNR) of MRI reconstruction by 3.1, 7.2, and 1.9 dB on three types of robustness
• Published in ICASSP 2023

SyncEdit | Go, gRPC, Protobuf
Mar. 2024 - Present
• Developing a client-server system that synchronizes edits on the same code file from geographically distributed programmers using Go and gRPC
• Leveraged vector logical clocks to determine invalid non-causal-related edits by different programmers

TaskTrack | React, Bootstrap, Axios, Spring Boot, Java, MongoDB
Dec. 2023 - Jan. 2024
• Developed the TaskTrack website that helps registered users create, update, track, delete tasks, and share tasks with their friends and teams
• Developed 3 types of web pages in 10 components using React.js and Bootstrap, and leveraged Axios to manage HTTP requests and responses
• Implemented RESTful APIs in 3 layers using Java, SpringBoot, and MongoDB to handle requests and manage the database, and leveraged JSON Web Token (JWT) and interceptor for user authentication and authorization
• Built unit tests of the backend APIs with Postman
[GitHub Repo]

Transfer-based Adversarial Attack with Few Queries (Undergrad thesis) | PyTorch, Wandb.ai
Jan. - June 2023
• Studied black-box adversarial attacks against deep learning models in the image classification task
• Reproduced 10 transfer-based and query-based attack approaches in PyTorch
• Proposed and Implemented LAQ-NI-FGSM and 2NI-FGSM that combined these two types of attacks
• Improved attack success rate by 7-15% with 15-20% fewer queries
[Abstract]

RoboMaster Autopilot | ROS2, Python
Sept - Oct. 2022
• Designed and developed a robot-host communication system using Robot Operating System (ROS) and Python, and migrated the computation-heavy workload to the host laptop
• Utilized infrared sensor and camera onboard the robot to detect obstacles and map the surroundings
• Leveraged D* lite routing algorithm to achieve robot auto-driving in an indoor playground
[Demo Video]

Jigsaw Puzzle Game Website | JavaScript, Paper.js
Oct. - Dec. 2021
• Built an online jigsaw puzzle game for playing in JavaScript with Paper.js graphics framework
• Created masks using polynomial interpolation to segment user-uploaded images into a vast number of tiles with respect to the chosen difficulty
• Bound tile moving and assembling functions to corresponding mouse events in Paper.js to smoothly respond to player's input
• Implemented 3 types of hints (highlighting, flashing, and rearranging) and a novel game mode, magnet mode, to facilitate puzzle-solving and improve user experience
[Gitee Repo (In Chinese)]

SMUG: Towards Robust MRI Reconstruction by Smoothed Unrolling
Hui Li, Jinghan Jia, Shijun Liang, Yuguang Yao, Saiprasad Ravishankar, Sijia Liu
International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 2023
[arXiv] [Code] [Poster]