Research Background
๐ป Undergraduate researcher in Software Engineering with a minor in Business Administration at South China University of Technology, Guangzhou. Focused on machine learning, knowledge-intensive systems, and data-centric applications, with experience spanning financial time-series modeling, scientific idea generation systems, and computer vision.
๐ฌ I pursue research that connects structured knowledge representations with machine learning, aiming to build transparent, rigorous systems for decision support and scientific reasoning. My recent work emphasizes multi-layer knowledge graphs, LLM-assisted expert agents, and data-driven modeling for complex domains.
Education
๐South China University of Technology, Guangzhou, China (Sep 2023 – Present)
B.Eng. in Software Engineering (Major) ยท Business Administration (Minor)
GPA: 3.61 / 4.00
Submitted Work
๐ [C.1] Wenbo Li, Xu Yingyu, Yiteng Chen, Peihao Chen, Lingwei Dang, Qingyao Wu, Huiping Zhuang, "Reconstructing Scientific Thinking: A Multi-Level Cognitive Engine for Idea-to-Blueprint Generation," submitted to the International Conference on Machine Learning (ICML) 2026.
Research & Projects
๐ Optimizing ETF Allocation with ML (SRP) (Mar 2025 – Present)
Tools: Distributed database, LSTM, Transformer, GNN
๐น Modeled financial time-series with a hybrid Transformer+LSTM and a risk-correlation GNN. Reported 60% annualized return and 19% cumulative return in backtests on 26 ETFs. Building an agent-based web system for strategy adjustment and drafting a research paper.
๐ง Reconstructing Scientific Thinking (AI4S) (Oct 2025 – Jan 2026)
Tools: Multi-Layer KG, LLM-Based Expert Agents, Cross-Layer Index Router
๐ Developed a 3-layer knowledge graph with cross-layer links for traceable scientific ideation. Built an LLM-driven index router and expert agents. Employed a 4-slot cognitive scaffold to enforce coherent reasoning and achieved 75% expert preference. โจ Expert + LLM evaluation showed systematic improvements across four cognitive dimensions.
๐ Search Image by Image (Mar 2025 – May 2025)
Tools: ResNet50, Faiss
๐ผ๏ธ Built a modular content-based image retrieval pipeline with three independent stages. Used ResNet50 for feature extraction and Faiss for indexing and retrieval, enabling an end-to-end workflow from batch processing to interactive visualization.
๐ค Reproduce Face Detection with MTCNN (Oct 2025 – Nov 2025)
Tools: PyTorch, OpenCV, TensorBoard
๐ฏ Reproduced the 3-network MTCNN cascade for face detection and alignment, achieving 99.1% mean precision on challenging benchmarks.
๐ Reproduce Neural Machine Translation with Seq2Seq (Oct 2025 – Nov 2025)
Tools: PyTorch, GRU, Attention, NLTK
๐ Implemented a Seq2Seq model with Bahdanau attention and optimized training to reach BLEU 28.45.
Skills
Programming Languages: ๐ Python, C, C++, โ Java, ๐ฆ Rust
Data Science & ML: ๐ฅ PyTorch, TensorFlow
Web Technologies: ๐ข Node.js, Vue
Database Systems: ๐๏ธ MySQL, SQL Server, TiDB
Research Tools: ๐ Overleaf, Origin, Zotero, ๐ค GPT, Prism, MATLAB
Other Tools: Git, ๐ง Linux, CVAT, Qt, JavaScript
Leadership & Service
๐ Class Representative, Software Engineering & Business Administration Dual-Degree Pilot Program, South China University of Technology (May 2024 – Dec 2025)
โ๏ธ Secretary Department Minister, SCUT IBM Club (May 2024 – May 2025)
๐ค Volunteer, Guangzhou Metro Star Volunteer Program (Apr 2025)
Recommended Reading
๐ Here are some books I recommend across various fields. Feel free to explore and share your own recommendations!