About Me
I am a senior undergraduate in the Yao Class (Institute for Interdisciplinary Information Sciences) at Tsinghua University, pursuing a B.S. in Computer Science and Technology (2023–2027, expected).
My research sits at the intersection of AI systems and computer architecture. I study the mechanisms that determine how efficiently large models run — from accelerator and memory architecture (e.g., 3D DRAM-based designs for language models), to LLM inference and serving (speculative decoding and draft-model design), to GPU kernel execution (persistent kernels).
Currently, I am a research intern at the Catalyst Lab, Carnegie Mellon University, working on persistent GPU kernels, and a research intern at Tencent, working on draft-model architectures for speculative decoding. Previously, I worked on LLM accelerator architecture at ArchipLab, Tsinghua University, and on search-based reasoning for Olympiad-style mathematics and physics problems at Tencent.
Research Interests
- AI accelerator architecture — memory-centric accelerator design for large language models, including 3D DRAM-based architectures.
- LLM inference & serving systems — speculative decoding, draft-model architectures, and serving efficiency.
- GPU systems & kernel execution — persistent kernels and system-level execution for GPU workloads.
- Agent systems & empirical LLM theory — tool-using agent workflows and measurement-driven studies of model behavior.
Education
- B.S. in Computer Science and Technology, Tsinghua University, 2023–2027 (expected). Institute for Interdisciplinary Information Sciences (Yao Class). GPA 3.8 / 4.0.
Selected Projects
Exploring memory-centric hardware for language model acceleration.
Architectures aimed at improving speculative token acceptance length.
Persistent GPU kernel research at CMU Catalyst Lab.
Experience
Academic
- Apr 2026 – PresentCatalyst Lab, Carnegie Mellon University — Research Intern. Persistent GPU kernels and system-level execution.
- Jan 2025 – PresentArchipLab, Tsinghua University — Research Assistant. LLM accelerator architecture with 3D DRAM.
Industry
- Jan 2026 – PresentTencent (WeChat Group) — Research Intern. Draft-model architecture for speculative decoding.
- Sep 2024 – Feb 2025ZhenFund — Investment Research Intern.
- Jul – Sep 2024Tencent (Machine Learning Platform) — Research Intern. MCTS-based reasoning for Olympiad-style tasks.
Blog
-
An introduction to this site and the research areas I plan to write about.
Miscs
- Board member of CPHOS, a nonprofit community of physics-competition alumni that organizes free competitions and academic seminars for high-school students (2023 – present).