About Me
Welcome to my blog! I'm Liam Fan (MELODIC-GIN), a passionate AI Full-Stack Engineer and Agent Architect.
👨💻 My Background
I specialize in architecting and developing tool-augmented LLM Agents. This blog is where I document my journey in building end-to-end AI applications, share insights on agentic workflows, and explore the intersection of modern web technologies and artificial intelligence.
🔧 Skills & Expertise
My focus is on building robust, scalable, and intelligent systems. Below is a snapshot of my technical capabilities and the core architectures I work with.
Technical Stack
| Domain | Technologies & Methodologies |
|---|---|
| AI & Agent Engineering | RAG, ReAct Agent Architecture, Workflow Orchestration, CoT Optimization, LangChain, Dify, OpenAI/Claude APIs, Gradio |
| Frontend Engineering | Next.js, React, TypeScript, Tailwind CSS (Utility-First), iOS(ObjC/Swift), Android, React Native, Flutter |
| Backend & Full-Stack | Node.js, Golang, Python (FastAPI/Flask), Vector DB (Milvus), SQL/NoSQL, HiveSQL |
| Platforms & Automation | Electron, Chrome Extensions, Playwright (RPA), Docker, CI/CD |
Core Architectures
Retrieval-Augmented Generation (RAG)
I architect RAG pipelines to ground LLMs in private, real-time data, mitigating hallucinations and enhancing response accuracy. The core principle is semantic search over a vector knowledge base.
The similarity between a query vector Q and document vectors D_i is calculated using Cosine Similarity:
ReAct Agentic Loop
For tasks requiring multi-step reasoning and interaction with external tools (APIs, databases), I design agents based on the ReAct (Reason + Act) paradigm. This framework enables the agent to create and adjust plans dynamically.
📬 Get In Touch
Feel free to reach out via email. I'd love to hear from you!