Skills for writing tilelang and debugging with CUDA toolkits.

A kernel library written in tilelang

An always-running agent, but is trustworthy and secure.

Control your claude code from your apple watch

金谷园饺子馆.Skill - 北邮旁的饺子馆

DFVG: A Heterogeneous Architecture for Speculative Decoding with Draft-on-FPGA and Verify-on-GPU.

TACIT — Tracked Agent Capabilities In Types

Make FP4 on 5090 Great Again

A Rust OS kernel autonomously implemented by Claude Code Opus/Sonnet.

Tree-structured context management for Claude Code via MCP server

RAFT contains fundamental widely-used algorithms and primitives for machine learning and information retrieval. The algorithms are CUDA-accelerated and form building blocks for more easily writing …

Check your Claude/Codex/Gemini, and learn how much money you save

Fast, small, and fully autonomous AI personal assistant infrastructure, ANY OS, ANY PLATFORM — deploy anywhere, swap anything 🦀

From Automated Idea Factory to Realization

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

A next.js web application that integrates AI capabilities with draw.io diagrams. This app allows you to create, modify, and enhance diagrams through natural language commands and AI-assisted visual…

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

Helpful kernel tutorials, examples and SKILLs for tile-based GPU programming

[ICML 2026]A framework to compare low-bit integer and float-point formats

The source of my blog.

GitHub Action to compile LaTeX documents

Open ABI and FFI for Machine Learning Systems

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

Fast Hadamard transform in CUDA, with a PyTorch interface