A skill for automatically optimizing CUDA code.

Scalable toolkit for efficient model reinforcement

A lightweight, AI-native training framework for large language models. Designed for fast iteration, reproducible experiments, and modular configuration across SFT, RLVR, and evaluation workflows.

AI agents running research on single-GPU nanochat training automatically

Project page of POET-X

Implementation for POET and POET-X for LLM pretraining

[KernelGYM & Dr. Kernel] A distributed GPU environment and a collection of RL training methods to support RL for Kernel Generations

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

Public repository for Agent Skills

Parmesan: Efficient Partitioning and Mapping Flow for DNN Training on General Device Topology

Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models

Spec-driven development (SDD) for AI coding assistants.

RLinf: Reinforcement Learning Infrastructure for Embodied and Agentic AI

SkillsBench evaluates how well skills work and how effective agents are at using them

[ICLR 2026 Blogpost Track Poster] JustRL: Scaling a 1.5B LLM with a Simple RL Recipe

Zhuofan Lin and Shiju Lin, "GPU-Accelerated Efficient Transduction for Logic Optimization", Design, Automation & Test in Europe Conference (DATE), 2026

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

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

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.

55nm CMOS Open Source PDK by ICsprout Integrated Circuit Co., Ltd.

PyTorch-native post-training at scale

A machine learning accelerator core designed for energy-efficient AI at the edge.

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

Adaptive Parallel PDF Parsing and Resource Scaling Engine

Fast and memory-efficient exact kmeans

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

verl: Volcano Engine Reinforcement Learning for LLMs

Domain-specific language designed to streamline the development of high-performance GPU/CPU/Accelerators kernels