Lei Chen
CV
Education, research interests, projects, and technical skills.
Lei Chen
Incoming M.S. Student in Artificial Intelligence
China · [email protected] · GitHub
Incoming M.S. student at UCAS / Institute of Computing Technology, interested in efficient deep learning, stereo vision, Rust, and large language models.
Education
University of Chinese Academy of Sciences / Institute of Computing Technology
M.S. in Artificial Intelligence, Beijing, China, 2026-present
- Incoming graduate student.
- Research direction: high-speed vision and multimodal information processing.
Southwest Jiaotong University
B.Eng. in Intelligent Manufacturing Engineering, Chengdu, China, 2026
- Undergraduate thesis: Optimization and Lightweight Deployment of Deep Learning-based Binocular Stereo Matching.
Research Interests
- Efficient Deep Learning and Model Compression
- Stereo Matching and 3D Vision
- Multimodal Perception
- Large Language Models and RAG
- Representation Learning and Embedding Models
- AI-powered Automation Tools
Projects
RAFT-Stereo Optimization and Lightweight Deployment
Optimized RAFT-Stereo for binocular stereo matching, focusing on refinement strategies, edge-aware loss, partial parameter unfreezing, and FP16 inference acceleration.
- Built experiments on the Middlebury2014 stereo dataset.
- Compared baseline, refinement-only training, refinement with edge-aware loss, and partial unfreezing.
- Reduced model size from about 43 MB to 22 MB with FP16 inference and improved inference speed from about 209.80 ms to 173.59 ms per image pair while maintaining comparable accuracy.
Competitive Price Monitoring System for E-commerce
Python and Playwright-based automation system for SKU-level price monitoring across Taobao/Tmall stores.
- Designed browser automation workflows for collecting product information and SKU-level pricing.
- Explored LLM-based SKU normalization and product model matching.
Skills
- Programming: Python, Rust, TypeScript, Shell
- Machine Learning: PyTorch, Computer Vision, Stereo Matching, Embedding Models, RAG, Model Compression, FP16 Inference
- Engineering: Playwright, Docker, Caddy, Cloudflare, GitHub Actions, Linux, Arch, CachyOS, self-hosted services
- Web / Tools: Hugo, Quartz, Chrome Extension Development
Languages
- Chinese: Native
- English: Working proficiency