A kernel library written in tilelang

A theoretical reconstruction of the Claude Mythos architecture, built from first principles using the available research literature.

Analyze computation-communication overlap in V3/R1.

My homework for Programming Practice.

A simple matrix calculator

Infiniband Verbs Performance Tests

Vibe Coding 指南 - 涵盖 Prompt 提示词、Skill 技能库、Workflow 工作流的 AI 编程工作站

CUDA Agent: Large-Scale Agentic RL for High-Performance CUDA Kernel Generation

A developer tool for disassembling, analyzing, debugging, and visualizing BPF object files.

NVIDIA NVSHMEM is a parallel programming interface for NVIDIA GPUs based on OpenSHMEM. NVSHMEM can significantly reduce multi-process communication and coordination overheads by allowing programmer…

Your own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞

UCCL is an efficient communication library for GPUs, covering collectives, P2P (e.g., KV cache transfer, RL weight transfer), and EP (e.g., GPU-driven)

Supercharge Your LLM with the Fastest KV Cache Layer

Accelerate local LLM inference and finetuning (LLaMA, Mistral, ChatGLM, Qwen, DeepSeek, Mixtral, Gemma, Phi, MiniCPM, Qwen-VL, MiniCPM-V, etc.) on Intel XPU (e.g., local PC with iGPU and NPU, discr…

TurboDiffusion: 100–200× Acceleration for Video Diffusion Models

Perplexity open source garden for inference technology

Mooncake is the serving platform for Kimi, a leading LLM service provided by Moonshot AI.

Tile primitives for speedy kernels

Official inference repo for FLUX.1 models

An early research stage expert-parallel load balancer for MoE models based on linear programming.

A guidance language for controlling large language models.

Disaggregated serving system for Large Language Models (LLMs).

High performance Transformer implementation in C++.