cv
Curriculum Vitae of Xinyi Zhang. Download the PDF version.
Contact Information
| Name | ZHANG Xinyi |
| Professional Title | Final-year PhD Candidate |
| csxyzhang@comp.hkbu.edu.hk | |
| Phone | +852-56027517 |
| Location | DLB 625G, 34 Renfrew Rd, HKBU, Kowloon Tong, Hong Kong SAR |
| Website | https://linekm.github.io |
Professional Summary
Final-year PhD candidate in the Department of Computer Science at Hong Kong Baptist University, with experience in designing and implementing high-performance, secure database systems and distribution-aware data structures.
Experience
-
2026 - Present Singapore
Visiting Scholar
Nanyang Technological University
Hosted by Dr. Qichen Wang.
- Working on “Learned Indexes for Query Optimization”, integrating learned index structures into structure-guided execution plans for acyclic queries.
-
2021 - Present Hong Kong SAR
PhD Candidate
Hong Kong Baptist University
Advised by Prof. Jianliang Xu.
- Designed novel algorithms and indexes for relational and vector databases to support efficient and secure query processing in a wide range of systems.
- Designed novel distribution-aware data structures to handle complex database workloads while ensuring robust performance.
- Research resulted in publications at top-tier conferences.
-
2020 - 2020 Beijing, China
Intern
ByteDance
- Optimized the inter-cluster communication for internal database clusters of ByteDance to enhance throughput and reduce latency.
-
2019 - 2020 Wuhan, China
Researcher
Huazhong University of Science and Technology
- Designed a fast consensus algorithm for blockchain combining SDN, dPOS and PBFT consensus algorithms.
- Built a practical blockchain system for ASTRI as an integrity-assured record for their public distributed database.
Education
Awards
-
2024 VLDB Student Travel Award
VLDB Endowment
-
2024 ACM SIGMOD Student Travel Award
ACM
-
2024 RPg Research Performance Award
Hong Kong Baptist University
-
2022 Excellent Teaching Assistant Performance Award
Hong Kong Baptist University
-
2018, 2021 Academic Excellence Scholarship
Huazhong University of Science and Technology
Publications
-
2026 HIRE: A Hybrid Learned Index for Robust and Efficient Performance under Mixed Workloads
Proceedings of the ACM on Management of Data (SIGMOD '26), vol. 4, no. 1
X. Zhang, L. Liang, A. Anastasia, and J. Xu. Full Paper.
-
2025 EDGE: DBMS-Empowered Boolean Decomposition for GIG Synthesis
ACM/IEEE Design Automation Conference (DAC '25), IEEE, pp. 1–7
R. Tang, X. Zhu, X. Zhang, L. Chen, X. Li, M. Yuan, and J. Xu. Full Paper.
-
2024 FedKNN: Secure Federated k-Nearest Neighbor Search
Proceedings of the ACM on Management of Data (SIGMOD '24), vol. 2, no. 1
X. Zhang, Q. Wang, C. Xu, Y. Peng, and J. Xu. Full Paper.
Skills
Languages
Interests
Projects
-
FedKNN: Secure Federated k-Nearest Neighbor Search (SIGMOD '24)
A system for secure and privacy-preserving federated kNN search, achieving up to 4.8× efficiency improvement over SOTA by optimizing local computations with a distribution-aware algorithm.
- Secure federated similarity search
- Distribution-aware local computation
- Up to 4.8× efficiency improvement over SOTA
-
HIRE: A Hybrid Learned Index for Robust and Efficient Performance (SIGMOD '26)
A hybrid in-memory index combining learned predictions with traditional robustness, which reduces tail latency by 98% and improves throughput by 41.7× compared to SOTA approaches.
- Hybrid learned + traditional index design
- 98% tail-latency reduction
- 41.7× throughput improvement
-
Privacy-Preserving Approximate Nearest Neighbor Search (Under Review)
A secure ANN index structure within trusted execution environments, mitigating side-channel attacks by enforcing oblivious memory access patterns on proximity graph indexes. Achieves plaintext-level throughput with provable privacy guarantees via a distribution-aware oblivious access algorithm.
- Oblivious proximity-graph access in TEE
- Plaintext-level throughput with provable privacy
- Distribution-aware oblivious algorithm
-
Authenticated Approximate Nearest Neighbor Search (Under Review)
VIDA, a novel scheme for authenticated graph-based ANN search that leverages vector commitments and Inner Product Arguments to verify distance computations, drastically reducing the overall data transmission size. MeST, a proof aggregation mechanism, compresses the verification of multiple vectors into a single cryptographic proof. Up to 5.2× speedup in end-to-end latency and 99.3% reduction in data transfer time.
- Vector commitments + Inner Product Arguments
- MeST proof aggregation
- Up to 5.2× end-to-end speedup; 99.3% data-transfer reduction
References
- Prof. Xu Jianliang — Head & Chair Professor, Hong Kong Baptist University
- Dr. Wang Qichen — Assistant Professor, Nanyang Technological University
- Prof. Byron Choi — Associate Head & Professor, Hong Kong Baptist University