Skip to content
View lyonsno's full-sized avatar
29 followers · 2 following

Achievements

Achievement: Pair Extraordinairex3Achievement: Pull Sharkx3Achievement: YOLOAchievement: QuickdrawAchievement: Arctic Code Vault Contributor

Achievements

Achievement: Pair Extraordinairex3Achievement: Pull Sharkx3Achievement: YOLOAchievement: QuickdrawAchievement: Arctic Code Vault Contributor

Highlights

  • Pro

Block or report lyonsno

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
lyonsno/README.md

Noah Lyons

I build GPU and ML systems for Apple Silicon and the browser: MLX kernels and model ports, browser-native WebGPU inference, and spatial tools for inspecting generated worlds.

Accepted upstream

See the full contribution index, selected reviews, and independent deployment evidence.

Browser-native GPU inference

  • @kaminos/webgpu-inference-kit is a composable WebGPU runtime for tensors, kernels, staged programs, route adapters, and cooperative scheduling.
  • moge-webgpu runs MoGe-2 depth and surface-normal estimation in browser-native compute shaders.
  • sharp-webgpu turns a single image into 3D Gaussian splats in the browser.
  • sf3d-webgpu generates textured 3D meshes in browser-native WebGPU.
  • kimodo-webgpu runs NVIDIA's 282M-parameter text-to-motion diffusion transformer in browser WebGPU, with DDIM sampling, denoising, and forward kinematics on the client.

Apple Silicon ML

  • mlx-nf4 is a standalone native Metal NF4 extension for stock MLX, verified from a clean source install on M2 Pro and exercised end to end by double-quantized GPT-2 and the 9.3B mlx-ideogram4 image pipeline.
  • cosmos3-mlx is a full-source MLX port of NVIDIA's 16B Cosmos 3 Nano world model for text/image-to-video and joint video/audio generation.
  • wilor-mlx rebuilds WiLoR hand-pose estimation end to end in MLX.
  • trellis2mlx and pixal3d-mlx bring image-to-3D generation to Apple Silicon without PyTorch.

Current work

Kaminos is very much a work in progress: it is the research bench where I explore the ideas that interest me most right now, especially browser-native ML, spatial interfaces, and tools for inspecting and shaping generated worlds. The WebGPU inference kit above is the most mature reusable component to emerge from that work.

Contact: n.lyons53@gmail.com

Pinned Loading

  1. wilor-mlx wilor-mlx Public

    WiLoR hand pose estimation for Apple Silicon — rebuilt end-to-end in MLX for real-time use

    Python

  2. mlx-ideogram4 mlx-ideogram4 Public

    Ideogram4 9.3B text-to-image on Apple Silicon via NF4 Metal kernels. 11.5 GB peak memory.

    Python 2 2

  3. pixal3d-mlx pixal3d-mlx Public

    MLX-native Pixal3D (SIGGRAPH 2026) image-to-3D inference for Apple Silicon. Pure MLX, no PyTorch.

    Python 1 1

  4. moge-webgpu moge-webgpu Public

    Single-image depth and surface normal estimation running entirely in the browser via WebGPU compute shaders. Port of MoGe-2 (ViT-Large + ConvStack) from PyTorch.

    JavaScript

  5. kaminos kaminos Public

    Kaminos (κάμινος) — hallucinated world kiln. WebGPU/Three.js workbench for meeting, creating and exploring locally generated virtual worlds.

    JavaScript

  6. mlx-nf4 mlx-nf4 Public

    NF4 / NormalFloat4 quantization kernels for MLX

    Python