Fast and memory-efficient classical machine learning operators

Compact and Agent-Native MoE Training System

A kernel library written in tilelang

The open-source agent-serving project

Our first fully AI generated deep learning system

A compact implementation of SGLang, designed to demystify the complexities of modern LLM serving systems.

Fast and memory-efficient exact attention

cuTile is a programming model for writing parallel kernels for NVIDIA GPUs

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

Perplexity open source garden for inference technology

Building the Virtuous Cycle for AI-driven LLM Systems

JAX support for tvm-ffi abi

Open ABI and FFI for Machine Learning Systems

Ship correct and fast LLM kernels to PyTorch

A Python-embedded DSL that makes it easy to write fast, scalable ML kernels with minimal boilerplate.

a size profiler for cuda binary

An extremely fast Python package and project manager, written in Rust.

VS Code extension for syntax highlighting C++/CUDA/HIP code in PyTorch load_inline() strings

RFC document, tooling and other content related to the array API standard

AGENTS.md — a simple, open format for guiding coding agents

TensorRT LLM provides users with an easy-to-use Python API to define Large Language Models (LLMs) and supports state-of-the-art optimizations to perform inference efficiently on NVIDIA GPUs. Tensor…

🎡 Build Python wheels for all the platforms with minimal configuration.

A next generation Python CMake adaptor and Python API for plugins

Tilus is a tile-level kernel programming language with explicit control over shared memory and registers.

Minimum example for deploying Apache TVM's Relax IR using C++ API

JaxPP is a library for JAX that enables flexible MPMD pipeline parallelism for large-scale LLM training

Distributed Compiler based on Triton for Parallel Systems

A Datacenter Scale Distributed Inference Serving Framework