Projects.

Tensors and Dynamic neural networks in Python with strong GPU acceleration

HarfBuzz text shaping engine

Lemonade helps users discover and run local AI apps by serving optimized LLMs right from their own GPUs and NPUs. Join our discord: https://discord.gg/5xXzkMu8Zk

cuda-oxide is an experimental Rust-to-CUDA compiler that lets you write (SIMT) GPU kernels in safe(ish), idiomatic Rust. It compiles standard Rust code directly to PTX — no DSLs, no foreign language bindings, just Rust.

cuTile Rust provides a safe, tile-based kernel programming DSL for the Rust programming language. It features a safe host-side API for passing tensors to asynchronously executed kernel functions.

Open ABI and FFI for Machine Learning Systems

Universal Virtualization Management Platform 可扩展的通用虚拟化管理平台，支持 Proxmox VE / LXD (GPU) / Incus (GPU) / Docker / Podman / Containerd / Qemu / Kubevirt

Open-source agent runtime — SSH-native isolation, eBPF egress policy, Kubernetes + LXC backends, GPU passthrough, MCP-native CLI

Pure Go FFI for WebGPU - CGO-free GPU access via wgpu-native bindings

Three.js-compatible application specializer with a Rust/Wasm + WebGPU execution core: compiles ordinary Three.js apps into specialized update kernels, packed GPU state, and persistent render schedules. Full r186 compatibility, zero feature cuts, honest 2x/3x benchmark gates.