Skip to content
View zianglih's full-sized avatar
🌰
🌰
60 followers · 54 following

Achievements

Achievement: Pair Extraordinairex2Achievement: Pull Sharkx2

Achievements

Achievement: Pair Extraordinairex2Achievement: Pull Sharkx2

Block or report zianglih

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
zianglih/README.md

Hi there 👋

I am 栗子昂, an MTS @ humans& ai working on the full-stack of LLM performance engineering. I graduated from University of Michigan, Ann Arbor (BS), after I transfered and spent 2 years at Chinese University of Hong Kong, Shenzhen.

Something I have been working on recently:

  • mxfp8 & nvfp4 RL
  • topk for sparse attention

Some of the places I previously worked/interned at:

  • NVIDIA GPU architecture simulation team
  • NVIDIA DevTech Compute team
  • Google Gemini GPU performance team
  • Samsung OpenCL compute team

I used to enjoy cycle-level GPU kernel optimization but I no longer consider it an important problem. My work has shfited more into low-precision numerics and model co-design.

Pinned Loading

  1. NVIDIA/TransformerEngine NVIDIA/TransformerEngine Public

    A library for accelerating Transformer models on NVIDIA GPUs, including using 8-bit and 4-bit floating point (FP8 and FP4) precision on Hopper, Ada and Blackwell GPUs, to provide better performance…

    Python 3.3k 718

  2. flashinfer-ai/flashinfer flashinfer-ai/flashinfer Public

    FlashInfer: Kernel Library for LLM Serving

    Python 5.6k 961

  3. radixark/miles radixark/miles Public

    Miles is an enterprise-facing reinforcement learning framework for LLM and VLM post-training, forked from and co-evolving with slime.

    Python 1.3k 190

  4. sgl-project/sglang sgl-project/sglang Public

    SGLang is a high-performance serving framework for large language models and multimodal models.

    Python 27.4k 5.8k