News
Serving as Chair for FRecSys ’26 — the 1st Frontier Recommender Systems Workshop, co-located with ACM SIGOPS ATC 2026.
Academic Papers
- HeteroGNN: A Heterogeneous Stage Division Based GNN Training Framework to Maximize CPU-GPU Parallelism ICME 2025
- Rearchitecting Buffered I/O in the Era of High-Bandwidth SSDs FAST 2026
- ASMA: An Anisotropy Scaling Memristor-based Accelerator for LLM Inference ICCD 2025
- Rethinking the Request-to-IO Transformation Process of File Systems for Full Utilization of High-Bandwidth SSDs FAST'25
- AIS: An Active Idleness I/O Scheduler to Reduce Buffer-Exhausted Degradation of Solid-State Drives ACM TACO
- RomeFS: A CXL-SSD Aware File System Exploiting Synergy of Memory-Block Dual Paths SoCC'24
- HEncode: A Highly Modularized and Efficient FPGA QC-LDPC Encoder Using High Level Synthesis ICCD'24
- A Study on Data-Layout Optimization in Memory for High-Performance Erasure Coding
- Analyzing performance degradation for wide stripe erasure codes
- PMLDS: An LSM-tree Direct Managed Storage for Key-value Stores on Byte-addressable Devices
- Leveraging Orbital Information and Atomic Feature in Deep Learning Model
Experience
Kuaishou Technology
Algorithm Engine Platform — Intern
Apr. 2026 – Present · Beijing
- Worked on the generative recommendation serving system, supporting online retrieval for the production deployments of OneRec and OneRec-V2.
Tencent
Advertising Engineering — Intern
May 2024 – Sep 2024 & Apr 2025 – Jun 2025 · Shenzhen
- int8 quantization in ad scoring system: >30% GPU kernel speedup, >2% system throughput gain.
- Added latency-analysis logs to ad recall system; tuned parameters to raise QPS from 10,000 to 15,000.
Shengquan Hengyuan Investment
Quantitative Finance Research — Intern
Jun 2023 – Aug 2023 · Nanjing
- >30% annual cumulative abnormal return using high- and low-frequency signals.
- Accelerated time-series and multi-factor model training with GPU servers.
Miracle Plus
Investment & Operations — Intern
May 2023 – Jun 2023 · Beijing
- Led diligence on a networking-industry startup that closed a Series A in the tens of millions.
SYSU AI4Science
Data Analyst
Jan 2022 – Jun 2022 · Guangzhou
- Predicted Synthetic Lethality using random-walk + graph convolution with multi-view fusion.
- Built message-passing models on CATL crystal database to predict physical properties.
Academic Projects
Model Optimization
- Profiled DGL conv bottlenecks with Nsight; optimized GCN normalization for a 30% speedup over DGL.
- C++/CUDA extensions for Python; CUDA operator overloading.
Reed-Solomon Code Optimization
- Compared wide- vs narrow-stripe erasure codes (ISA-L / Jerasure); reached 128 Gbps single-node throughput via OpenMP.
SSD Optimization
- Built SSDTEST to characterize performance degradation; developed I/O scheduler to control tail latency.
Education
University of Hong Kong
PhD, Computer Science
2025 – present · adv. Prof. Yiming Qiu
Huazhong University of Science and Technology
MSc, Computer Science
2022 – 2025 · Outstanding Graduate (Top 20%)
Sun Yat-sen University
BSc, Computer Science
2018 – 2022
Talks & Presentations
- ICME'25, IEEE International Conference on Multimedia & Expo 2025, Nantes, France. Website
- FAST'25, 23rd USENIX Conference on File and Storage Technologies, Santa Clara, California. Video
- SoCC'24, 15th ACM Symposium on Cloud Computing, Redmond, Washington.
- HiPEAC'25, 20th High Performance, Edge And Cloud computing, Barcelona, Spain.
Certificates & Awards
Awards
- Sangfor Scholarship (Apr. 2024)
- HUST Scholarship, Second Prize (Oct. 2023)
- Distinguished Activist in Community Affairs (Oct. 2023)
- SYSU Scholarship, Second Prize (Oct. 2021)
- 2019 ACM-ICPC SYSU Competition, Second Prize (Oct. 2019)
Certificates
- IELTS 7.0 · GRE 320 · TOEFL 106