I'm currently completing my MSc in Computer Science at University College Cork (UCC), Ireland.
My thesis work explores how AI, geospatial data, and modular architecture can be combined to create intelligent, exploratory systems.
- I work in English, and also speak Mandarin and Cantonese — occasionally I even code in all three 😉.
I’m actively looking for a project I can meaningfully contribute to. (Or if you want to build any airports, stadiums or hospitals I could also help you with that. I am a former architect.)
- Thesis-based personal demo. Subject to continuous development. Source is available, but not open for contribution.
This project adopts an RAG pipeline to experiment how LLMs, urban landmark metadata, and player movement can be integrated into a spatially-aware, riddle-based exploration system.
- Java Spring Boot backend for session, puzzle, and player logic
- Python-based microservice for real-time LLM riddle generation (GPT / llama.cpp local deployment)
- Leaflet.js frontend for map interaction, player orientation, and target logic (MVP Stage)
- Wikipedia/OSM-based landmark "metadata" preprocessing pipeline
- MongoDB for persistent storage of riddles, sessions, and landmark data
Future:
- Epistemic planning agent module under construction (to adapt riddles based on player knowledge state)
- Containerization via Docker/Kubernetes and cloud hosting (AWS)
🔗 Source code
🔗 Riddle Agent
🔗 Landmark Processor
🧾 Thesis Abstract (Available on request)
Languages & Platforms
Java 17 · Python 3.10 · JavaScript · HTML/CSS · Bash
Frameworks
Spring Boot 3.x · Flask · Leaflet.js · LLM API (OpenAI / llama.cpp)
Data & Infrastructure
MongoDB · RESTful API · Docker · Kubernetes & Jenkins Pipeline (planned)
Tech Focus Areas
LLM Integration · Geospatial Interaction · Prompt Engineering · Microservice Design · Urban Data Processing · Agent Behavior Modeling
Project Overview: Developed a comprehensive autonomous driving path planning system based on Hybrid A* Algorithm and Model Predictive Control (MPC), implementing a two-tier architecture with Global Planner and Local Planner
- Core Technologies:
- Global Planner: Implemented Hybrid A* algorithm for vehicle motion planning with non-holonomic constraints
- Local Planner: Real-time trajectory tracking control using MPC with kinematic bicycle model
- Collision Detection: Real-time obstacle detection and avoidance algorithms
- Visualization Interface: Real-time simulation interface built with PySide6 and PyQtGraph
Tech Stack: Python 3.12+, NumPy, SciPy, CVXPY, PySide6, OpenCV, Matplotlib, PyQtGraph
DevOps-4-DevOps is a pluggable multi-application DevOps console Dashboard system. MVP aims at building itself from source code through the entire pipeline: build → test → package → deploy ("deploy itself" as MVP). The extension will also integrate/orchestrate the CICD pipeline of other Apps (locally or on cloud). Frontend uses React (with charts), CI/CD uses Jenkins, containerization uses Docker.
I'm especially interested in:
- Agentic systems & epistemic logic
- Narrative-based urban experience
- Web3-based open collaboration frameworks
📬 Always open to connect with:
- Builders working on creative infrastructure
- Web3 explorers & DAO practitioners
- Game/interaction designers with a spatial focus