On AI Lounge since Nov 2026
Joe Zhuo
Member #4,061DeveloperL2Student, Independent
Waterloo CE | AI Inference
How karma works. Everyone starts with 5. You get +1 each time another member upvotes one of your posts or comments (your own upvotes don't count), and lose it if they take the upvote back. Post what you're building, answer questions and help other members to earn more.
Achievements
Plugged In ×2Connected sources: GitHub, Hugging Face, LinkedIn, X, website.How: Connected 2 sources (GitHub, LinkedIn, X or a website) in Settings
Plugged In×2Featured
ShipperProjects on your profile.How: Added 1 project to their profile
Shipper
No posts yet
Joe Zhuo hasn't posted yet.
Pinned projects
An always-on-top Windows desktop companion built with Tauri 2 and Rust featuring an animated pixel crab that monitors Claude Code sessions, tracks activity, and handles approvals.
Claude Code skills that delegate light/medium tasks to Codex CLI and Kiro CLI
A single-GPU LLM inference engine in C++20 and CUDA. It serves TinyLlama-1.1B-Chat-v1.0 in BF16 with FP32 accumulation, matches HuggingFace output token for token, and serves many concurrent requests with a paged KV cache, a paged decode attention kernel, continuous batching, custom fused kernels, sampling on the GPU, and CUDA Graphs.
inferenceinference-enginellmmachine-learningsystems-programming
A Claude Code plugin that stops Claude from over-explaining.
About
Joe Zhuo is a Computer Engineering student at the University of Waterloo who works independently as a developer on AI inference systems. Joe built crawde, an always-on-top Windows desktop companion written in Tauri 2 and Rust that watches Claude Code sessions, and stop-yap, a Claude Code plugin that reduces verbose replies. Joe's focus is LLM-Inference-Engine, a single-GPU engine in C++20 and CUDA that serves TinyLlama with a paged KV cache, continuous batching, and CUDA Graphs.
