Skip to content
View jameswei's full-sized avatar

Highlights

  • Pro

Block or report jameswei

Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
jameswei/README.md

Hi, I'm James

I'm a builder, engineer, and developer. I like turning fuzzy ideas into working systems: AI agents, developer tools, infrastructure, and small utilities that make daily engineering feel lighter.

Current Workbench

AI and learning-first systems

  • tiny-rag-lab - a learning-first RAG engine for classic RAG scenarios.
  • tiny-duo-infer - a tiny inference engine with Apple Silicon and NVIDIA GPU backends.
  • tiny-gpt - GPT built from scratch through hands-on ML practice.
  • neetcode-submissions - algorithm and ML practice notes from NeetCode.

Agents and developer workflow

  • human-in-loop-workflow - a simple, well-behaved workflow for human-reviewed agent work.
  • zot - a lightweight Go coding-agent harness fork where I contributed session behavior fixes upstream.
  • zot-goal - a TypeScript extension for goal-oriented agent runs.
  • holon - a local workbench for agents doing continuous work across sessions and commands.

Practical tools and personal apps

  • porta - a macOS menu-bar app for finding and killing orphan listening ports.
  • climb-coach - a climbing coach app and a personal gift project.
  • Windows_Hotkey - useful scripts and hotkeys for working on Windows.

Earlier backend and infrastructure projects

Open Source Contributions

Recent merged contributions:

Long-Running Interests

  • Distributed systems, databases, messaging, and backend infrastructure
  • Go, Python, Java, and the occasional shell script
  • RAG systems, local inference, coding agents, and developer experience
  • Learning by rebuilding small versions of useful systems

Pinned Loading

  1. tiny-gpt tiny-gpt Public

    GPT built from scratch — assembled from my NeetCode ML course submissions

    Python 1

  2. tiny-duo-infer tiny-duo-infer Public

    A tiny inference engine implementation for learning purpose. Supports duo backends: Apple Silicon and Nvidia GPU backend.

    Python 1

  3. climb-coach climb-coach Public

    Climb coach, a gift for my son -- Lele

    TypeScript 1

  4. holon-run/holon holon-run/holon Public

    A local workbench for agents doing continuous work across sessions, commands, human input, and external events.

    Rust 130 12

  5. tiny-rag-lab tiny-rag-lab Public

    a learning-first RAG engine for classic RAG scenario.

    Python 2

  6. patriceckhart/zot patriceckhart/zot Public

    Yet another coding agent harness, lightweight and written in go.

    Go 282 23