Logic-First Developer & AI Orchestrator
Year 4 CS Student @UG-CS-HKBU | Developer Intern @BytewiseHK
I don't just write code; I orchestrate agents.
I specialize in Hierarchical AI Workflows—using "Manager" agents for architecture/state
and "Worker" agents for implementation, preventing context decay while moving fast.
The context window is the new bottleneck. Here is how I manage it:
- Manager-Worker Pattern: I maintain a "Manager" chat session that holds only the Project Spec & State (zero code). It delegates tasks to ephemeral "Worker" sessions.
- Spec-Driven: The "Manager" outputs strict requirements; "Workers" implement them. This separates logic from syntax.
- Verification Gates: AI writes the implementation, but I verify the logic before merging.
- Hybrid Retrieval: Combining Vector Search (RAG) with Knowledge Graphs to ground the agents (as seen in my FYP, TCM-Sage).
| Project | Description | Stack |
|---|---|---|
| TCM-Sage | FYP: An LLM-augmented diagnosis assistant for Traditional Chinese Medicine. | RAG · Knowledge Graphs · Streamlit · Multi-LLM |
| @BytewiseHK | Developing chatbot features and agentic workflows using various LLMs. | Pinecone · n8n · LLMs (Claude/GPT/Gemini/etc.) |
| Freelance | E-commerce solutions (e.g., Verona Gelato) and newsletter automation. | Vue 3 · Supabase · Vercel · Railway |
Agentic IDEs & Design
AI Models & Apps
Production Stack
- Agentic Workflows: Refining the "Manager-Worker" pattern for larger teams.
- Hybrid Retrieval: Improving RAG accuracy by combining vector similarity with graph traversal.
- Product Thinking: Moving beyond "does it run?" to "does it solve the user's problem?"
| Topic | What We Can Talk About |
|---|---|
| AI Workflows | How to use Antigravity/Stitch/LLMs to speed up development cycles. |
| Context Management | Techniques to keep LLMs "smart" over long projects (Hierarchy, RAG, Summarization). |
| Formula 1 | The drama of the 2021 season or just vibing on race weekends. |
| Cars | Porsches (911 is the dream), JDM legends, and car photography. |
I am currently focused on my FYP and Internship, but I am open to lightweight discussions or brainstorming on:
- RAG & Knowledge Graph architectures.
- New AI-Assisted development tools (Antigravity, Stitch, Spec-kit, etc.).
- 🐹 TWICE fan (Sana is my bias).
- 🏋️♂️ Currently trying to build wider shoulders at the gym.
- 🏎️ Dream Garage: An indecisive mix of a Porsche 911 GT3 RS, a comfy RV for road trips, and maybe a daily driver... I can't decide.