Overview#
I am currently a senior software engineer working at AMD ROCm (previously at Huawei Ascend), building vLLM inference engine for GPU/NPU software ecosystem (focusing on multi-modality inference, structured output and OOT (Out-Of-Tree) hardware extensibility). Before this, I was a student at Beijing Jiao Tong University (BSc/MSc), majoring in communication engineering. You can see my projects at Github, or see my posts at Zhihu.
Experiences#
AMD
Senior SDE
June '26 - Present
Working at AMD ROCm, building vLLM inference engine for AMD GPU software ecosystem.Huawei
Software Engineer
August '24 - June '26
Working at Huawei Ascend, building vLLM inference engine for Ascend NPU software ecosystem.Huawei
Software Engineer
July '23 - August '24
Working at Huawei Quality and Process IT Department, building MetaERP BI (Business Intelligence) for PSI (purchase-sale-inventory) IT system.Beijing Jiao Tong University
Master
September '21 - June '23
Studying at School of Electronic and Information Engineering, focusing on NAS layer security of wireless communication.Beijing Jiao Tong University
Bachelor
September '16 - June '20
Studying at School of Electronic and Information Engineering, majoring in communication engineering.
Open Source Contributions#
Overall, I have contributed 152 PRs to the vLLM ecosystem, focusing on multi-modal inference, structured output and OOT hardware extensibility. Find more details about all PRs I have contributed here.
- Outside collaborator of vllm:
A high-throughput and memory-efficient inference and serving engine for LLMs
- Core contributor to vllm-ascend:
Community maintained hardware plugin for vLLM on Ascend
Projects#
This repo is used for archiving my notes, codes and materials of cs learning.
A curated collection of Claude Code agent skills that accelerate the entire vLLM development lifecycle.
Programming Skills#
Multilingual Skills#
- English: IELTS overall band score of 6.5
Contact me#
Feel free to drop me an email, gmail or qq-mail are both available.
You can also have my WeChat through the picture shown below:

NOTE: Remember to add a description about your intentions before sending a request.