Google's Engineering Practices documentation

Ultra-low-latency, high-throughput multiprocess transport over SHM and mmap. LMAX-Disruptor-style cross-process ring substrate.

Cross-architecture CUDA kernels for SVDQuant (W4A4 with low-rank correction)

LLM model quantization (compression) toolkit with HW acceleration support for Nvidia, AMD, Intel GPU and Intel/AMD/Apple CPU via HF, vLLM, and SGLang.

Compile docker images into a single self-contained binary

Tools for converting ACPI DSDT to Device Tree Source for CIX Sky1 boards

Model compression toolkit engineered for enhanced usability, comprehensiveness, and efficiency.

A Python DSL to write Nvidia PTX for Hopper and Blackwell in JAX and PyTorch

The headless browser for AI agents and web scraping

Open-source AI sandbox infrastructure with unified API for VMMs -- Firecracker, QEMU and libkrun.

Secure and fast microVMs for serverless computing.

Dynamic per-token early exit for LLM inference. Skip layers tokens don't need

rvLLM: High-performance LLM inference in Rust. Drop-in vLLM replacement.

🤖FFPA: Extends FlashAttention-2 via Split-D for large headdims, 1.5x~3×↑🎉 vs SDPA, up to 430T🎉 on H200.

Qwen3-TTS is an open-source series of TTS models developed by the Qwen team at Alibaba Cloud, supporting stable, expressive, and streaming speech generation, free-form voice design, and vivid voice…

DeepGEMM: clean and efficient FP8 GEMM kernels with fine-grained scaling

Enable true multi gpu capability in Comfy UI using XDiT XFuser and FSDP managed by Ray

OBLITERATE THE CHAINS THAT BIND YOU

[ICML 2026] Jacobi Forcing: Fast and Accurate Diffusion-style Decoding

The open-source agent-serving project

Fast and accurate AI powered file content types detection

AR 3D object detection for iPhone with LiDAR — YOLO 2D + BoxerNet 3D lifting

🎨 NeMo Data Designer: Generate high-quality synthetic data from scratch or from seed data.

OpenShell is the safe, private runtime for autonomous AI agents.