Hi, I'm Huang

I build things for the web, for language learners, and for fun. Here are the projects I work on — I also write notes about what I learn along the way.

OctoCounts preview

OctoCounts

octocounts.com

A free SLOC counter for public GitHub repositories. GitHub shows language bars but not actual line counts — OctoCounts fills that gap. Paste a repo URL and the Rust backend (Axum + Tokio) downloads the archive tarball, runs tokei across 200+ languages, and caches the report by commit SHA so repeat queries are instant. Also available as a Chrome / Edge / Firefox extension that adds a SLOC card directly to GitHub repo sidebars, with JSON / plain-text / PNG-badge export.

EchoPod preview

EchoPod

echopod.clothpath.com

A language-learning app built around AI subtitles. EchoPod turns your favorite podcasts into interactive bilingual study material: subtitles and playback stay in sync, and guided echoing sessions train your listening and speaking. Available on the iOS App Store, with an Android alpha program.

Ketsuin 結印 preview

Ketsuin 結印

ketsuin.clothpath.com

A web-based input method that lets you type with ninja hand signs. Form one of the 12 classic seals in front of your webcam — a YOLOX-Nano model running on ONNX Runtime WebAssembly detects the sign entirely in your browser (no data leaves your device), and a T9 predictive engine maps sign sequences to text. Includes a challenge mode where you race hand-seal sequences against a global leaderboard.

QwenASR preview

QwenASR

github.com/huanglizhuo/QwenASR

A fast, pure-Rust, CPU-only inference engine for Qwen3-ASR speech-to-text, tuned for Apple Silicon with hand-written NEON / Accelerate / AMX-aware kernels — no Python, no GPU, no tensor framework, only libc. It transcribes a 28-second clip in 613 ms on an M5 Pro (46× realtime), beating GPU-based MLX implementations. Supports offline and streaming transcription, live capture with VAD, SRT/VTT subtitles, structured JSON, and forced alignment. Install with cargo install qwen-asr-cli.