I'm
Chuqing Gao,
a CS PhD student.
Sounds like chew-ching 😋🪙, if that makes it easier for you.
a CS PhD student.
Sounds like chew-ching 😋🪙, if that makes it easier for you.
Work with Prof. Jianguo Wang at Purdue University. Focus on database for LLM, database optimization and large-scale data solutions.
Earned my bachelor's degree in CS at University of Science and Technology of China, where I'm fortunate to work with Prof. Cheng Li.
Worked as an intern of the System Research Group in Microsoft Research Asia, collaborating with Dr. Ying Cao on deep learning compiler.
SIGMOD'24 Revisiting B-tree Compression: An Experimental Study.
This benchmark revisits and evaluates 7 B-tree compression techniques, offering a systematic comparison of their effectiveness in modern database environments. By analyzing space efficiency, query performance, and insertion trade-offs, our study provides practical guidance for optimizing B-tree implementations in real-world systems.
VLDB Journal'24 Optimizing LSM-based Indexes for Disaggregated Memory.
This extended work on dLSM introduces improvement in near-data compaction for disaggregated memory systems. An adaptive compaction strategy is proposed to dynamically optimize the distribution of compaction tasks between compute and memory nodes based on hardware configurations and real-time workload conditions.
Deep Learning Compiler Optimization (2021)
Conducted a case study on LSTM and contributed to the development of a deep learning compiler system, focusing on operator fusion, memory optimization, and data flow efficiency for frameworks like PyTorch and TensorFlow. The system, FractalTensor, was later published on SOSP'24.
SQL Query Plan Optimization (2020)
We developed a heuristic algorithm for optimizing left-deep-join tree in SQL query plan and redesigned the compute flow, achieving obvious performance improvement on the given workload on a commercial open-source database engine. This project, as part of the Openlookeng performance optimization track, win the national championship.
Graph DataBase File System (2019)
We utilize graph database Neo4j to build a FUSE-based file system with a fantastic web UI. The files are connected in the graph based on their contents. We used some machine learning techniques to extract keywords and description. Due to the limitation of early AI techniques, GDBFS can only process some simple files.
