Skip to content
View mayooot's full-sized avatar
🤔
Woodstock jamming
🤔
Woodstock jamming
35 followers · 22 following
  • Beijing

Achievements

Achievement: QuickdrawAchievement: Pair ExtraordinaireAchievement: Pull Sharkx2Achievement: Starstruck

Achievements

Achievement: QuickdrawAchievement: Pair ExtraordinaireAchievement: Pull Sharkx2Achievement: Starstruck

Block or report mayooot

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
mayooot/README.md

Hi, I'm Ming 👋

I build AI infrastructure on Kubernetes — currently working on multi-billion-scale image vector retrieval (Milvus / FAISS) and AI platform (training & inference) at Megvii.

What I work on

  • ☸️ Kubernetes internals: operators, GPU scheduling (HAMi / Volcano / Kueue / Device Plugin)
  • 🔍 Vector search at scale: index selection, recall/latency/memory benchmarking, capacity planning
  • 🖥️ GPU infrastructure: NCCL tuning, DCU/NPU enablement, container-level GPU management — earlier built a multi-tenant GPU cloud (A100/H800, per-second billing) serving external users
  • 🏗️ Multi-tenant K8s clusters (VCluster) — previously at DataCanvas
  • 📊 Measuring before believing — most of my work starts with a benchmark

Upstream contributions kubernetes/kubernetes · golang/go · apple/containerization · vcluster · aibrix · HAMi · kueue · all PRs →

Writing: Zhihu column (Chinese)

📫 harrymingh@gmail.com

Pinned Loading

  1. gpu-docker-api gpu-docker-api Public

    Easier than K8s to lift and lower the gpu number of docker container and scale capacity size of volume.

    Go 82 15

  2. detect-gpu detect-gpu Public

    Detect-GPU is an http server that detect the host for NVIDIA GPU info.

    Go 21 1

  3. build-nccl-tests-with-pytorch build-nccl-tests-with-pytorch Public

    Build NCCL-Tests and configure SSHD in PyTorch container to help you test NCCL faster!

    Cuda 13

  4. alfred-json-format alfred-json-format Public

    🤘An Alfred workflow that formatting a copied json string to the pasteboard using jq.

    14 2