Join arcSYSuJoin arcSYSu

arcSYSu (ARChitecture and SYStem Upscaling @ SYSU), also known as ARX, is a research group led by Xianwei. The name also forms a recursive acronym: "arcSYSu refines computing system uses." We focus on GPU systems, compilers, HPC/AI infrastructure, and system software frameworks. Members are students directly advised (or co-advised) by Xianwei. arcSYSu is also part of the Interdisciplinary Research Center (xRC), with close collaboration across groups in projects, publications, and mentoring. ☆ Note: arcSYSu is an independent research group, not a formal organizational unit.

We work on computer systems, GPU architectures, and compilers, with an emphasis on optimizing systems for HPC and modern AI workloads (LLMs, agents, etc.). Our typical workflow is, Application → System Design → Back to Application: Analyze workloads and identify bottlenecks; Design system/software/hardware optimizations; Implement and validate improvements. We also explore HPC applications and AI workloads, often spanning both systems and algorithms. Typical tools and stack: profilers (nvprof/rocprof), LLVM/Clang, CUDA/ROCm, containers/K8s, and LLM frameworks (e.g., vLLM, SGLang). Common languages: C/C++, Python, Bash, CUDA/HIP/OpenCL. ☆ Current focus areas: GPU systems, compilers, ML systems, HPC platforms, and application acceleration.

Not exactly. While our work is close to hardware, we primarily take a systems/software approach to optimize hardware–software co-design. We emphasize a full-stack perspective (algorithm → software → hardware), but typically do not involve circuit-level design. Representative conferences to follow include: SC, ASPLOS, HPCA, ISCA, MICRO, PPoPP, SOSP, OSDI, CGO, NeurIPS, AAAI, ICLR.

Self-motivated and disciplined; A drive for excellence, beyond degree requirements; Interest in systems and real-world applications; Solid CS fundamentals and proficiency in 1–2 programming languages; Good fit and communication style (mutual). ☆ Prior experience with our specific tools is NOT required.

Experience across industry, academia, and management organizations; Emphasis on practical, long-term, and well-rounded training; Hands-on mentoring throughout the research process; Access to collaborations and strong recommendation channels; Strong platform with state-of-the-art resources and projects. ☆ Note: these can also be downsides depending on your expectations—fit matters.

Topic selection: aligned with projects, background, and interests; Research: structured guidance with weekly 1-on-1 / group discussions; Skills: hands-on ability, independent thinking, communication, teamwork; Exposure: encouraged to attend conferences and engage externally. I follow a collaborative advisor–advisee model, where: we work side-by-side, learn from each other, and make discoveries together. (See references from Harcol@CMU and Herman@ColumbiaU for more details.)

Systems research is inherently full-stack, so we aim for both: - Depth: expertise in a focused area - Breadth: end-to-end understanding (algorithm → system) There are no hard publication quotas, but students are expected to achieve substantial growth in: research, engineering, communication, and overall professionalism. Your work should translate into strong, visible outcomes (projects, papers, systems), enabling competitiveness in both academia and industry. ☆ Note: the expectations are NOT that easy to reach—please consider carefully.

Internships are strongly encouraged, especially when aligned with your research. We see industry and academia as complementary, not competing: industry provides scale and real-world exposure; academia provides depth and rigor. Both paths offer excellent opportunities—this is ultimately a personal choice. Be Prepared, Be Open.

Graduates with full-stack systems expertise have broad options, including: - System/architecture roles (e.g., Huawei, AMD, NVIDIA); - Software/hardware engineering in leading tech companies (e.g., ByteDance, Tencent, Alibaba, Google); - Startups in HPC/AI infrastructure and acceleration.

~24 current PhD/MS students (2021–2025), expected steady at ~20; Most students have internship experience (NVIDIA, Intel, Alibaba, ByteDance, Tencent, Kimi, Huawei, etc.); 8 MS graduates so far, placed at Intel, HKUST(GZ), NVIDIA, Huawei, MetaX, JD, etc.; ~1/3 pursue PhD (directly or after MS); A few undergraduate researchers each year. See more at: https://arcsysu.tech/

Yes—motivated undergraduates are welcome. That said: Prioritize broad fundamentals and full-stack skills; Avoid locking into a narrow direction too early; Treat lab participation as an extension of coursework. To officially join (as UG/RA), you are expected to: Actively participate for at least one semester; Engage in meetings, discussions, or projects; Be recognized by the group.

Email: zhangxw79@mail.sysu.edu.cn We can have a coffee chat in person or have virtual meeting. You are also encouraged to contact current students via the website: https://arcsysu.tech/people.