With over 25 years of experience, I transform complex hardware concepts into stable, production-ready networking products. I thrive at the intersection of Hardware, Software, and FPGA — especially during the critical Board Bring-up (Day 0) and system integration phases.

Beyond my professional work, I'm also passionate about AI, web technologies, and developer tools — exploring how modern software can make development smarter and more accessible.

• Embedded Linux: Kernel development, U-Boot, Driver optimization, RootFS (Yocto/Buildroot)
• Board Bring-up & HW/SW Integration: Expert on NXP (LX2160), Marvell (ARM), and Intel platforms
• Security: Secure Boot implementation (Cisco Trust Anchor), 5G-ORAN (Genevisio)
• High-Speed Interfaces: 25G SerDes, PTP/SyncE optimization
• Networking Products: Full lifecycle delivery — concept → prototyping → mass production → field support
• Cross-Team Collaboration: Bridging gaps between SW, HW, and FPGA teams; mentoring through code reviews and architectural guidance

Notable background: Delivered mission-critical networking equipment at Cisco (ASR1K, ISR, Network Modules).

What I build in my spare time — exploring new technologies and creating tools that help developers work smarter.

Real board bring-up workflow for M5Stack CoreS3 — ESP-IDF, UART Console, sensor diagnostics & power management

• Built with: C, ESP-IDF framework, Clean Architecture
• Highlights: Real-time sensor monitoring, Interactive UART command console, Power management & thermal monitoring, Modular & extensible diagnostic modules

An AI round-table in your browser — ask one question, get answers from 4 AI models with an unbiased Chairperson summary. Features anonymous members, randomized order, deep-dive one-on-one chats, and an Offline Demo Mode that needs zero API keys.

• Built with: Pure frontend HTML/CSS/JS (single file, no backend)
• Highlights: Anti-bias architecture, recursive summarization, customizable system prompts

A private, self-hosted AI stack (inference, chat, RAG, and MCP tool calling) that keeps models and documents inside your own network boundary — with real cost, tokens/s, and failure data recorded instead of assumed.

• Built with: Docker Compose, Ollama, Open WebUI, MCP, cloudflared, Bash scripts
• Highlights: Private by default (Ollama never published, outbound-only tunnel), local stack and RAG + MCP verified end to end, measured vs. projected figures explicitly labelled, 76 dated engineering decisions documented, bilingual docs
• Status: Local stack and RAG + MCP working; GPU endpoint and agent runtime still in progress

A full-stack web platform showcasing popular GitHub projects with AI-powered features, including trending project scraping, smart search, and AI-generated recommendations

• Built with: AI-assisted development (TypeScript, React, Python)
• Features: Smart search, AI recommendations, trending scraper, responsive UI

Job market analytics and scraping tool for employment market insights

• Built with: Python (AI-assisted)
• Purpose: Employment market data analysis

28.9M param offline LLM on M5Stack CoreS3 — bilingual story generation with Google PLE, ~9.5 tok/s

• Built with: C (ESP-IDF/Arduino), Python (training pipeline)
• Highlights: 14.9MB quantized model in Flash, ~9.5 tok/s inference, temperature sampling, Chinese LCD rendering via M5GFX

Exploring AI applications and business automation concepts

• Built with: Python (AI-assisted)

Collection of Claude Code configuration templates for different project types

• Built with: Shell scripts, configuration management

• Advanced AI integration patterns
• Full-stack web development
• Cloud deployment strategies
• Modern embedded tooling and automation

• Open-source AI projects
• GitHub ecosystem tools
• Web scraping and data analysis
• Developer productivity tools
• Embedded systems tooling

• 🤖 Artificial Intelligence & Machine Learning
• 🛠️ Embedded Systems & Hardware-Software Integration
• 🔐 Secure Boot & Platform Security
• 🌐 Web Development & Full-Stack Applications
• 📊 Data Analysis & Web Scraping
• 💼 Business Automation

• LinkedIn: ian-chang-56136479
• GitHub: @ian0318git
• Location: Melbourne, Australia
• Open to: Senior Embedded Systems roles, collaboration on interesting projects

• 🎯 I believe in the power of AI to democratize software development
• 🚀 I don't just write code — I identify systemic bottlenecks and bridge communication gaps between cross-functional teams
• 🌍 Always excited to explore new technologies and frameworks
• 💪 Love contributing to open-source and learning from the community
• 📖 Passionate about mentoring and raising team standards through robust code reviews
• 🔄 Seen products through the entire lifecycle — from concept to mass production and field support

擁有 超過 25 年 經驗，我擅長將複雜的硬體概念轉化為穩定、可量產的網路產品。我尤其熱衷於 硬體、軟體與 FPGA 的交匯處，特別是在關鍵的 Board Bring-up (Day 0) 和系統整合階段。

在專業工作之外，我也對 AI、Web 技術和開發者工具 充滿熱情 — 探索現代軟體如何讓開發更智慧、更容易。

• Embedded Linux: Kernel 開發、U-Boot、驅動程式優化、RootFS (Yocto/Buildroot)
• Board Bring-up 與 HW/SW 整合: 精通 NXP (LX2160)、Marvell (ARM) 和 Intel 平台
• 安全機制: Secure Boot 實作 (Cisco Trust Anchor)、5G-ORAN (Genevisio)
• 高速介面: 25G SerDes、PTP/SyncE 優化
• 網路產品: 完整生命週期交付 — 概念 → 原型 → 量產 → 現場支援
• 跨團隊協作: 橋接 SW、HW 與 FPGA 團隊之間的溝通差距，透過 Code Review 與架構指導提升團隊水準

重要經歷: 在 Cisco 參與關鍵網路設備開發 (ASR1K、ISR、Network Modules)。

業餘時間的探索 — 嘗試新技術，建立幫助開發者提高效率的工具。

M5Stack CoreS3 真實 Board Bring-up 工作流程 — ESP-IDF、UART 控制台、感測器診斷與電源管理

• 使用方式: C, ESP-IDF 框架, Clean Architecture
• 亮點: 即時感測器監控、互動式 UART 命令控制台、電源管理與溫度監控、模組化可擴展診斷模組

私有、自架的 AI 技術棧（推論、對話、RAG 與 MCP 工具呼叫），讓模型與文件留在自己的網路邊界內，並以實測記錄成本、tokens/s 與失敗案例，而非憑空假設。

• 使用方式: Docker Compose、Ollama、Open WebUI、MCP、cloudflared、Bash 腳本
• 亮點: 預設私有（Ollama 不對外發布、僅用對外連出的 tunnel）、本地堆疊與 RAG + MCP 已端到端驗證、實測與推估數據明確標註、76 筆有日期的工程決策紀錄、中英雙語文件
• 狀態: 本地堆疊與 RAG + MCP 可運作；GPU endpoint 與 agent runtime 仍在進行中

在瀏覽器中開 AI 圓桌會議 — 問一個問題，4 個 AI 同時回答，再由匿名主席做出總結。特色包括匿名化委員、隨機打亂順序、一對一深聊、以及免 API Key 的離線模擬模式。

• 使用方式: 純前端 HTML/CSS/JS（單一檔案，無後端）
• 亮點: 反偏頗架構、遞迴總結、可自訂 System Prompt

一個全面的平台，用於發現和探索 GitHub 上最受歡迎的開源項目，由 AI 建議和實時趨勢數據提供支持。

• 使用方式: AI 輔助開發（TypeScript、React、Python）
• 功能: 智能搜尋、AI 建議、趨勢爬蟲、響應式 UI

職位市場分析和爬蟲工具

• 使用方式: Python（AI 輔助）
• 用途: 就業市場數據分析

28.9M 參數離線 LLM 在 M5Stack CoreS3 上 — Google PLE 雙語故事生成，~9.5 tok/s

• 使用方式: C（ESP-IDF/Arduino）、Python（訓練管線）
• 亮點: 14.9MB 量化模型存入 Flash、~9.5 tok/s 推論、隨機溫度取樣、M5GFX 中文字型 LCD 顯示

探索 AI 應用和業務自動化概念

• 使用方式: Python（AI 輔助）

針對不同項目類型的 Claude Code 配置模板集合

• 使用方式: Shell 腳本、配置管理

• 高級 AI 集成模式
• 全棧 Web 開發
• 雲端部署策略
• 現代嵌入式工具與自動化

• 開源 AI 項目
• GitHub 生態工具
• 網頁爬蟲和數據分析
• 開發者生產力工具
• 嵌入式系統工具鏈

• 🤖 人工智慧與機器學習
• 🛠️ 嵌入式系統與軟硬體整合
• 🔐 Secure Boot 與平台安全
• 🌐 Web 開發與全棧應用
• 📊 數據分析與網頁爬蟲
• 💼 業務自動化

• LinkedIn: ian-chang-56136479
• GitHub: @ian0318git
• 地點: 澳洲墨爾本
• 開放機會: Senior Embedded Systems 職位，也歡迎有趣的項目合作

• 🎯 我相信 AI 的力量能夠民主化軟體開發
• 🚀 我不只是寫程式 — 我善於識別系統性瓶頸，並在跨職能團隊之間搭建溝通橋樑
• 🌍 對探索新技術和框架充滿熱情
• 💪 喜歡為開源做貢獻並向社區學習
• 📖 熱衷於透過 Code Review 和架構指導提升團隊標準
• 🔄 經歷過產品從概念到量產再到現場支援的完整生命週期