itamaker - Overview

Hi, I'm Zhaoyang 👋

💼 Staff AI/ML Engineer • 🤖 AI Agents & LLM Platforms • ✈️ Open to International Relocation

Shipping small experiments, learning in public, and keeping software a little playful.

What I'm Doing

• Building practical AI systems and developer tools.
• Turning day-to-day automation patterns into installable agent skills.
• Exploring better workflows across Golang, Python, and LLM applications.
• Shipping infrastructure-friendly products with Docker and Kubernetes.

Current Projects

• 🐱 kitten-tts-go - Go implementation of KittenTTS, a lightweight ONNX text-to-speech engine.
• 🔐 gauth2apple - Split a Google Authenticator export QR into individual otpauth:// QR codes for Apple Passwords.
• 🧩 skills - An index for my standalone agent skills.
• 🗂️ go-workspace-skills - A standalone skill for configurable Go workspaces.
• ✨ stitch-skills - A standalone skill for Google Stitch workflows.
• 💻 workspace - A Makefile-driven workspace for syncing and managing the my project repos.
• 🚀 go-chrome-ai - A cross-platform Chrome profile patcher written in Go, with both CLI and GUI modes.
• 🍺 homebrew-tap - A Homebrew tap for casks maintained and shipped under the my namespace.
• 🧠 delta - A deep learning based end-to-end natural language and speech processing platform.
• 💬 dstc9 - DSTC9 dialog-system challenge code, experiments, and related research work.
• ⚒️ skillforge - Draft and scaffold OpenClaw-ready skills from briefs, tool catalogs, and JSON specs.
• 🔎 runlens - Analyze agent and tool JSONL traces for failures, flaky paths, latency outliers, and token usage.
• 📚 ragcheck - Evaluate retrieval and RAG runs with Precision@k, Recall@k, MAP, nDCG, and offline answer judges.
• 🃏 promptdeck - Render, batch, and optimize prompt templates from JSON variables, experiment matrices, and score files.
• 🧪 datasetlint - Audit JSONL datasets for duplicates, semantic leakage, label conflicts, and other train/eval quality issues.
• ⚙️ dotfiles - A chezmoi-managed personal dotfiles repo for Neovim, Starship, and reproducible local setup.

GitHub Activity

Connect

Make it useful. Keep it light. Ship it.