Guanghua Li

I am Guanghua Li, a Ph.D. student in Data Science and Analytics at The Hong Kong University of Science and Technology (Guangzhou), advised by Prof. Qiong Luo. I expect to graduate in January 2027.

I build GPU-accelerated query processing systems for graph queries and multi-vector retrieval. My research focuses on turning irregular, data-intensive query workloads into efficient parallel execution while keeping systems programmable and scalable to large datasets. My work spans tensor-based graph query processing, subgraph queries beyond GPU memory, and late-interaction retrieval for text and multimodal data.

Looking further ahead, my long-term interest is in data systems for the AI era—how data should be managed, retrieved, and processed to provide reliable and efficient infrastructure for AI applications.

Previously, I was a research assistant in the Database Research Group at The Chinese University of Hong Kong, advised by Prof. Jeffrey Xu Yu. I received my bachelor’s degree in Computer Science and Technology from Wuhan University in 2022, where I was advised by Prof. Yuanyuan Zhu.

Research

• VectorHit — ongoing research on GPU-accelerated multi-vector retrieval.
• Vora — a vector-based subgraph query engine that uses a single GPU to process graphs much larger than device memory. Accepted at ICDE 2027.
• TenGraph — a graph query engine built from PyTorch tensor operations, with one codebase for multicore CPUs and GPUs. PVLDB 2024.
• GCSM-BU — GPU continuous subgraph matching for batch updates, using a query-oriented formulation to avoid duplicate matches. APWeb 2026.

For research inquiries, email gli945@connect.hkust-gz.edu.cn. See my CV for education and research experience.