Scalable toolkit for efficient model reinforcement

slime is an LLM post-training framework for RL Scaling.

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

htop-like TUI for real-time RDMA network monitoring.

Can LLMs Write Correct and Efficient GPU Communication Code?

The end of web parsing. The beginning of scalable pixel-native search. link: https://pixelrag.ai/

mKernel: fast multi-node, multi-GPU fused kernels

Ring attention implementation with flash attention

Fast and memory-efficient exact kmeans

A Proof-oriented Programming Language

A Distributed Attention Towards Linear Scalability for Ultra-Long Context, Heterogeneous Data Training

A benchmark of real-world DL kernel problems

Share your GPU without MIG or MPS

Google Workspace CLI — one command-line tool for Drive, Gmail, Calendar, Sheets, Docs, Chat, Admin, and more. Dynamically built from Google Discovery Service. Includes AI agent skills.

Automated High-Performance GPU Kernel Generation

Fast and Furious AMD Kernels

CUDA-L1: Improving CUDA Optimization via Contrastive Reinforcement Learning

DeepGEMM: clean and efficient BLAS kernel library on GPU

Mirage Persistent Kernel: Compiling LLMs into a MegaKernel

Research works from Tencent AI Lab regarding self-evolving agents

SysMoBench: Evaluating AI on Formally Modeling Complex Real-World Systems

Autonomous GPU Kernel Generation & Optimization via Deep Agents

Building the Virtuous Cycle for AI-driven LLM Systems

A fast communication-overlapping library for tensor/expert parallelism on GPUs.

Tile-based language built for AI computation across all scales

Tile-Based Runtime for Ultra-Low-Latency LLM Inference