Xiaoyu Zhang
Xiaoyu Zhang (BBuf) is an individual open-source AI-infrastructure developer and technical creator, not a company. Public profiles describe his focus as LLM inference, CUDA kernels and AI infrastructure; he works at RadixArk, contributes to SGLang, and publishes through the GiantPandaCV/GiantPandaLLM presence.
At a Glance
- AI-infrastructure engineers and researchers
- Teams deploying LLMs and multimodal models
- SGLang, vLLM, TensorRT-LLM and TokenSpeed developers
- GPU/kernel performance engineers
- +1 more
AI Tools by Xiaoyu Zhang
(1)
AI-Infra-Auto-Driven-SKILLS
Agent Skills for LLM Serving
Discussions
No discussions yet
Be the first to start a discussion about Xiaoyu Zhang
Latest News
AI-Infra-Auto-Driven-SKILLS repository updated with expanded SGLang explainer-video workflow and plugin version 0.9.2
AI-Infra-Auto-Driven-SKILLS v0.1.5 released with model Day-0 support, profiler layer guides and 118 bilingual model histories
HPC-Ops × SGLang integration article acknowledges Xiaoyu Zhang (BBuf) for backend, review and design contributions
LMSYS publishes Agent-Assisted SGLang Development, featuring BBuf's AI-Infra-Auto-Driven-SKILLS and KDA-Pilot workflows
Products & Services
Personal publishing identity and website associated with Xiaoyu Zhang, used for AI/ML engineering articles and projects. The GitHub profile links www.giantpandacv.com and describes him as creator of the GiantPandaLLM official account.
Open-source agent-ready playbooks and tools for framework-neutral LLM serving benchmarks, capacity planning, Torch Profiler analysis, pipeline analysis, compute simulation, SGLang model Day-0 support, humanized review, production incident triage, model-architecture diagrams and SGLang explainer videos. The plugin metadata credits BBuf as author.
A public BBuf project applying kernel-design-agent-style workflows to SGLang and AI-infrastructure kernel optimization; the July 2026 LMSYS article describes its public B200 diffusion summary and SGLang integration work.
The BBuf GitHub profile lists how-to-optim-algorithm-in-cuda among selected systems, alongside SGLang and AI-infrastructure projects, reflecting practical CUDA/kernel engineering education and experimentation.
Market Position
Zhang's public work is positioned as open-source, evidence-driven infrastructure engineering rather than a commercial hosted product. AI-Infra-Auto-Driven-SKILLS differentiates through cross-framework comparison, reproducible benchmark contracts, profiler/source attribution and SGLang-specific operational knowledge; the adjacent competitive ecosystem includes vLLM, TensorRT-LLM and other inference-optimization/tooling projects.
Leadership
Founders
Xiaoyu Zhang (BBuf)
GitHub identifies Xiaoyu Zhang as BBuf, working at RadixArk and as the creator of the GiantPandaLLM official account. His public profile README describes him as a core developer associated with SGLang/RadixArk focused on LLM inference, CUDA/CUTLASS/Triton kernels and AI-infrastructure systems; his Hugging Face profile links the same BBuf identity and RadixArk organization.
Executive Team
Xiaoyu Zhang (BBuf)
Independent open-source developer; Core Developer associated with SGLang; RadixArk technical staff
GitHub describes him as working at RadixArk and creator of the GiantPandaLLM official account. His public technical profile centers on LLM inference, CUDA kernels, fast serving and AI infrastructure.
Founding Story
The subject is a person rather than an incorporated venture. His public work evolved around sharing computer-vision/AI engineering knowledge through GiantPandaCV/GiantPandaLLM and building open-source tooling for practical model serving, profiling, kernel optimization and reproducible AI-infrastructure workflows.
Target Markets
- AI-infrastructure engineers and researchers
- Teams deploying LLMs and multimodal models
- SGLang, vLLM, TensorRT-LLM and TokenSpeed developers
- GPU/kernel performance engineers
- Open-source contributors and technical learners
- LLM and multimodal-model serving deployment and benchmarking
- GPU inference performance debugging and Torch Profiler analysis
- SGLang model onboarding and upstream contribution planning
- CUDA kernel engineering and inference optimization
- AI-infrastructure incident response and reproducible engineering workflows
- Technical education and open-source AI/ML knowledge sharing