University of California San Diego
La Jolla, CaliforniaMaster of Science in Computer Engineering
UCSD M.S. in Computer Engineering
Yang Zhang
Building practical AI and embedded systems with strong engineering depth.
I am Yang Zhang, a computer engineering graduate student at UC San Diego with a background in communication engineering, embedded systems, computer vision, and machine learning.
About
Profile
My work sits at the intersection of AI systems and hardware-aware engineering. I have built image synthesis pipelines, optimized real-time vehicle detection models, and developed embedded control systems for robotics and autonomous platforms.
I am especially interested in computer vision, efficient deep learning, embedded intelligence, and bringing research ideas into reliable implementations.
Education
Academic Background
University of Electronic Science and Technology of China
Chengdu, ChinaBachelor of Engineering in Communication Engineering
- GPA: 3.8/4.0, ranked in the top 10% of the cohort.
- Relevant coursework: AI and Machine Learning, Communication Networks, Embedded Processors, Signal and Systems, Circuit Analysis, Information Security.
- Honors include UESTC Merit Scholarship for Exceptional Student and Hainan Free Trade Port Merit Scholarship.
Research
Lab Experience
State Key Laboratory of Computer Architecture
Research Assistant, State Key Lab of Processor
- Designed embedded systems tailored for CMOS image sensors to improve image processing quality and output performance.
- Developed and optimized image signal processing algorithms in Python, MATLAB, and C.
- Implemented functions such as noise reduction, sharpening, and color correction.
Center for Advanced Computer System
Research Assistant
- Analyzed Maximum Information Coefficient data to uncover relationships across datasets.
- Improved the MIC algorithm for stronger correlation discovery and clearer visualization.
- Optimized the processing workflow to raise both accuracy and efficiency.
Projects
Selected Work
Generative AI
Image Synthesis from Text Prompts
Built an anime-style image synthesis pipeline using FLUX.1 with LoRA fine-tuning and mixed-precision training under hardware constraints.
- Improved Inception Score by 29.6% compared to Stable Diffusion v1.5.
- Reduced FID by 36.5% and validated gains with multi-dimensional preference testing.
Computer Vision
Vehicle Detection with MobileNetV4 + RT-DETR
Optimized RT-DETR for real-time vehicle detection in constrained environments by integrating a lightweight MobileNetV4 backbone.
- Cut parameter count by 65% and computational complexity by 75% versus RT-DETR-L.
- Validated performance on COCO-format vehicle datasets.
Embedded Robotics
AI Acupuncture Robot
Developed a robot arm control system based on STM32F4 with PID control, sensor feedback, Wi-Fi communication, and OLED status display.
Optimization + ML
Thermistor Network Parameter Optimization
Combined genetic algorithms and neural networks to optimize thermistor network parameters with higher precision and lower computational cost.
Autonomous Systems
Embedded Vehicle Navigation Platform
Led the vision recognition subsystem using OpenMV and Python for robust autonomous navigation and task execution.
Publication
Recent Publication
Improved RT-DETR Based on MobileNetV4 for Vehicle Detection
Published in the 2025 8th International Conference on Advanced Algorithms and Control Engineering (ICAACE).
Contact