Resume

Master of Science, Computer Science

University of California - Los Angeles, Los Angeles, CA, USA

• GPA: 4.0
• Teaching: Teaching Assistant for CS174A: Intro to Computer Graphics. Leading the discussion session with more than 50 students, lecturing advanced graphic knowledge and Javascript.
• Teaching: Teaching Assistant for CS33: Introduction to Computer Organization. Leading the discussion session with more than 50 students, lecturing basically computer organization and architecture knowledge.
• Core Courses: Cloud Computing, Data Mining, Adversarial Attack and Robustness, Advanced Computer Architecture, Hardware For Machine Learning.

Bacholar of Science, Computer Science

University of California - Los Angeles, Los Angeles, CA, USA

• GPA: 4.0
• Honor: Summa Cum Laude, Dean's Honors List
• Teaching: Learning Assistant for CS32: Intro to Computer Science, lectured more than 50 students.
• Core Courses: Algorithms, Computer Organization, Operating Systems, Computer Architecture, Network, Programming Languages, Database, Software Engineering, Computer Vision, NLP, Machine Learning

Vision Algorithm Research Engineer

Kuaishou Technology

• Designed and developed vision-language multimodal models with 0.6B parameters.
• Implemented SOTA algorithms to generate descriptions for un-classified magic effects with Pytorch.
• Leveraged pre-trained models and fine-tuning techniques to reduce human-labeled data usage by 60%

Student Researcher

UCLA Prof. Bolei Zhou's Group

• · Developed an improved Human-in-the-loop Reinforcement Learning (RL) method. The trained agent can master driving tasks in less than 30 minutes on a home PC, saving more than 90% of the training time compared to traditional RL methods. Using PyTorch. Paper accepted to NeurIPS 2023 Spotlight.
• Built and tested a platform for large-scale traffic scenario modeling and simulation for RL, IL, and autonomous driving. Paper accepted to NeurIPS 2023.
• Implemented a more photorealistic simulation environment for training RL autopilot agents using UE4.

Student Researcher

UCLA Structure-Computer Interaction Lab

• Built a 2D LiDAR robotic navigation algorithm for a road identification system and improved the navigation accuracy by 30% (compared with multi-Ransac) for the robot while maintaining a low cost.
• Deployed road identification system on low-cost autonomous weed-control robot, using C++ and ROS.
• Created an inverse learning approach to generate the physical parameters (such as diameter) for the soft robot and increase the efficiency of collecting the parameters by 90%, using TensorFlow.

Student Researcher

UCLA Center for Neurobehavioral Genetics

• Processed RNA sequence data and performed PCA, data visualization and other data analysis on it, using Python and R.
• Proved that Low-coverage RNA sequencing is an effective approach in Expression quantitative trait loci studies and can save the cost by more than 50%. Learn more details via the paper Powerful eQTL mapping through low-coverage RNA sequencing.

AR Glass Assistants APP (Java, AR)

• Developed an AR Glasses Auxiliary Andriod App using Java, enabling image and sound capture and ensuring a s eamless user experience through efficient compression techniques.
• Integrating Google AR core to enable object detection and sound classification.
• Implemented a robust queuing system with congestion control to facilitate seamless data transfer between AR glasses and smartphones.
• Minimized latency effects through a special design, enhancing the real-time responsiveness of the system.
• Tested its resilience and reliability in real-world conditions for AR glasses.

Accelaration Library on Apple M Chip and NVIDIA GPU (MetalAPI, CUDA)

• Developed a high-performance GEMM operation program that can achieve 6 TFLOPS performance. Using C++ and Metal API.
• Developed an optimized Conv-2d program with CUDA, achieving comparable performance with cuBLAS

Gradient-based adversarial attacks against text transformers (Python, NLP)

• Using BARTScore and BLEURT as similarity constraints to perform a gradient-based attack against commonly- used text transformers to explore their potential robustness problems, using Python

C++-based Web Server (C++, CI/CD, Rest API)

• Built a NGINX standard web server with REST API capabilities using C++.
• Constructed a CI/CD Pipeline on GCP with Detailed log info, test coverage monitoring using Google Test, and codereview using gerrit. Setup a monitor dashboard to record up-time and request latency.
• Source Code

WebGL-based game (Javascript, WebGL, Computer Graphics)

• Designed and implemented an online basketball-shooting game with advanced graphic features, including shadow, texture, and reflections, using JavaScript and WebGL.
• Demo

Egglendar Online Calendar (React, SQL)

• Developed an online student calendar specially for international students, with features of importing and converting calendars, adding school course schedules, and finding classmates.
• Using Material-UI for designing and React for the front-end framework, and MySQL as the database.

• Dean's Honors List * 9
• Summa Cum Laude