I am a Ph.D. student in the Department of Computer Science and Engineering at the Hong Kong University of Science and Technology (HKUST), advised by Prof. Binhang Yuan and Prof. Wentao Zhang. I am currently a Qingyun Program Research Intern with Tencent Hunyuan’s Foundation Model Department. My research focuses on LLM data infrastructure and high-performance distributed systems for large-scale LLM data preparation, including data-operator management systems designed for ease of use, distribution, and reproducibility, as well as high-performance data processing built on Ray and Apache Spark. My earlier work includes data preparation systems for data-centric AI and image forensics, especially image manipulation detection and localization.
Before starting my Ph.D., I was a Research Assistant at Peking University from January 2025 to January 2026, where I worked with Prof. Wentao Zhang on data preparation systems for large language models. From July 2023 to June 2024, I was a full-time Research Assistant at Sichuan University, where I worked with Prof. Jizhe Zhou on image manipulation detection.
I received my B.Eng. in Computer Science and Technology from the College of Computer Science at Sichuan University in June 2023. My undergraduate thesis was recognized by the university as an Outstanding Undergraduate Thesis.
A composable data-pipeline and workflow system for LLM data preparation, spanning data generation, cleaning, evaluation, and orchestration.
Core contributor across pipeline compilation and execution, operator and registry engineering, storage, serving, and CLI tooling. I also built the RayOrch integration for transparent Ray data parallelism, in-memory storage, resource cleanup, and end-to-end tests.
A modular training and evaluation codebase and comprehensive benchmark for image manipulation detection and localization.
Joint first author, codebase designer, and coding lead. I designed the modular training and evaluation framework and have maintained its metrics, model zoo, CLI, packaging, releases, and tests.
A Ray-powered acceleration layer that turns MinerU PDF-to-Markdown parsing into a scalable multi-GPU and multi-node data pipeline.
Project initiator and core developer. I built the initial package and release workflow, then refactored the sequential stages into an asynchronous pipeline with cross-stage overlap, while maintaining benchmarks and compatibility.
Lightweight orchestration utilities for asynchronous Ray pipelines, with RayModule, overlapped microbatch execution, and DAG scheduling.
Project initiator and core developer. I built the initial orchestration abstractions and multi-model examples, refined RayModule and the pipeline executors, and integrated RayOrch into DataFlow for parallel operator execution.
Reviewer: AAAI 2027, NeurIPS 2026, BMVC 2026, ECCV 2026, CVPR 2026, ICLR 2026, AISTATS 2026, AAAI 2025, NeurIPS 2025, ICCV 2025, ICML 2025, CVPR 2025, ICLR 2025, AISTATS 2025, NeurIPS 2024, and ACM MM 2024.
Ph.D in Computer Science and Engineering, 2028 (expected)
HKUST
B.Eng. in Computer Science, 2023
Sichuan University
