Quansheng GU

古权胜

Building toward impactful AI infrastructure.

I am passionate about high-performance computing and AI infrastructure, currently exploring research opportunities and industry internships, aiming to build impactful systems for efficient inference. FLOPS IS ALL YOU NEED.

Email GitHub

Current Work

Sparse-vLLM

with Jitai Hao

Project

Sparse-vLLM

A sparse-first inference engine that rethinks KV cache layout, controller flow, and kernels, while also including DeltaKV compressor training and evaluation tooling.

Education

Harbin Institute of Technology, Shenzhen · Computer Science and Technology

School

Harbin Institute of Technology, Shenzhen

Major

Computer Science and Technology

Status

Sophomore

Campus

Campus Experience

HPC team work and technical exploration around AI systems.

HITSZ HPC Team Core Member

Dedicated to HPC and supercomputing competitions represented by ASC.

Special Experience

The 15th National Games Volunteer

Awarded Outstanding Volunteer for event service.

Interests

AI infraEfficient inference enginesHPC systems

Competitions

Competition Experience

Supercomputing and ICT competition records.

ASC

2026 ASC Student Supercomputer Challenge

A student supercomputing contest covering cluster deployment, parallel performance tuning, and HPC application optimization.

First Prize

Huawei ICT

Huawei ICT Competition Challenge Track

A Huawei-style ASC challenge on Ascend and Kunpeng heterogeneous computing, CANN operator development, and AI workload acceleration.

Third Prize

Tech Stack

Technical Stack

Languages, accelerators, frameworks, profiling tools, and HPC systems.

Languages

PythonC++

Hardware

NVIDIA GPUAscend NPUKunpeng CPU

Frameworks

PyTorchTensorRTCUDA Graphs

Tools / Systems

Nsight SystemsPyTorch ProfilerApptainerMPISlurm

Contact

gqs060905@outlook.com

GitHub

kuma-loong