⏰ Highlights: Our work on benchmarking materialized view query rewriting pipeline is accepted to VLDB 2026!
- Resume
- Email: xha102@sfu.ca
- I am currently a CS PhD candidate at Simon Fraser University, Canada.
- Actively looking for internships and collaborations. Please see my research interests below.
- I received my bachelor's degree from the Southern University of Science and Technology (SUSTech 南方科技大学), China.
- I worked as a product manager and a data engineer at Sohu (搜狐), China for three years. I designed and built the engine and data-governance system for the advertising business.
- Next-generation materialized view (MV) service with LLM:
- Step 1: Before applying LLM, I really want to figure out the performance of existing MV service. Our work [VLDB 2026] shows that MV-driven query rewriting and acceleration is far from being solved, as it is really a matter of "pipeline" (view enumeration, view selection, view-based query rewriting) while previous literatures only carried out single stage evaluation.
- Step 2: We have already seen the great potential of using LLM on this task. I am actively incorporating LLM as a semantic reasoner to view enumeration and view-based query rewriting.
- AI infrastructure: AI4DB tasks scheduling on GPU.
- Extensive AI4DB tasks require using GPU: Learned optimizers, LLM-powered UDFs, semantic queries etc., These tasks vary a lot on latency sensitivity, parametre size. Finding out an optimal solution to orchestrate them is my focus.
[VLDB 2026]: Benchmarking the Full Pipeline of Materialized-View-Based Query Rewriting (Xinjie Hu, Zhengjie Miao)