Projects

AI Fashion Assistant

RAG and LLM-Powered Fashion ChatBot leveraging Retrieval-Augmented Generation (RAG) to provide styling advice and outfit recommendations.

Live Demo Available

Application Demo

AI Fashion Assistant Demo Screenshot

Fashion chatbot interface

Implementation Overview

  • •Implemented RAG architecture with vector databases for fashion knowledge retrieval and semantic search
  • •Integrated Large Language Models for natural conversation generation
  • •Developed responsive React frontend with real-time chat interface and product integration
  • •Deployed backend using FastAPI with RESTful architecture

Key Achievements

  • ✓Deployed AI chatbot

Technologies Used

PythonFastAPILLMRAGVector DBReactRESTful API

Diversity-Driven News Recommender System

Contributed to Diversity-Driven news recommendation research and Informfully Recommenders development.

Published Research

Framework Architecture

Informfully Recommenders Framework Architecture

Source: Informfully Recommenders -- Reproducibility Framework for Diversity-aware Intra-session Recommendations. ACM RecSys 2025

Implementation Overview

  • •Developed a light-weight D-RDW (Diversity-driven Random Walk) recommendation model using Normalized Target Distribution (NTD) for dynamic multi-dimensional distribution control
  • •Developed re-ranking methods for increasing recommendation diversity
  • •Built open-source news text feature augmentation pipelines for NLP tasks (named entity extraction, classification, tokenization, transformers, sentiment analysis)

Key Achievements

  • ✓Published in ACM RecSys 2025
  • ✓Informfully Recommenders: Open-source framework developed

Technologies Used

PythonTensorflowPytorchMachine LearningNLPInformation Retrieval

Virtual Clothing Simulation & Fashion Applications

MPhil research project in Hong Kong focusing on computer-aided fashion design and virtual garment fitting systems

MPhil Research Project

Implementation Overview

  • •Developed an Intelligent Virtual Fitting system that automatically fits clothing patterns on customized 3D human body models
  • •Created algorithms for automatic pattern preprocessing and prepositioning, reducing setup time from hours to minutes
  • •Implemented garment simulation combining Finite Element Method (FEM) with real-time rendering for web-based performance
  • •Integrated personalized human body modeling techniques moving beyond standard mannequins to individual avatars

Key Achievements

  • ✓Achieved virtual try-on efficiency improvement: reduced garment simulation setup from hours to minutes
  • ✓Gained insights into streamlining fashion supply chain from manufacturing to retail

Technologies Used

Computer GraphicsFinite Element Method (FEM)3D ModelingVirtual RealityFashion TechnologyHuman Body ModelingReal-time Rendering

Drawing & Guessing: Interactive Web Game

Lab Course Project - Real-time multiplayer drawing and guessing game with collaborative gameplay and modern web technologies.

Production Deployed

Application Demo

Drawing & Guessing Game Demo Screenshot

Real-time multiplayer drawing and guessing gameplay interface

Implementation Overview

  • •Led team of 5 developers using Agile methodology and GitHub workflows
  • •Built responsive React frontend with HTML5 Canvas for drawing
  • •Developed RESTful API with Spring Boot and JPA for game management

Key Achievements

  • ✓Team leadership: Coordinated 5-person development team
  • ✓Deployed a game application supporting multiple players

Technologies Used

ReactSpring BootHTML5 CanvasJPAREST API

Chinese News Classification System

Course Project - Machine learning approaches for automated classification of Chinese news articles.

Completed Project

Classification Methodology

Chinese News Classification Methodology Overview

Comparative analysis of machine learning approaches for Chinese text classification

Implementation Overview

  • •Implemented and compared multiple classification approaches: Naive Bayes, SVM, Logistic Regression
  • •Developed Chinese text preprocessing pipeline including segmentation, stopword removal, and feature extraction
  • •Selected dataset and conducted experiments to evaluate classification accuracy

Key Achievements

  • ✓Implemented efficient Chinese text preprocessing pipeline
  • ✓Achieved over 80% accuracy in news classification task

Technologies Used

PythonScikit-learnNLTKJiebaChinese Text ProcessingTF-IDFSupervised Machine Learning

Blockchain Platforms Analysis & Comparison

A systematic analysis and comparison of major blockchain platforms, examining their consensus mechanisms, governance models, and technical architectures as part of a collaborative course project.

Completed Project

Implementation Overview

  • •Analyzed multiple blockchain platforms including Ethereum, Bitcoin, Binance Smart Chain, and emerging protocols
  • •Compared consensus mechanisms (Proof of Work, Proof of Stake, Delegated Proof of Stake) and their trade-offs
  • •Evaluated governance structures, tokenomics, and decentralization levels across different platforms
  • •Investigated scalability solutions, smart contract capabilities, and ecosystem development

Key Achievements

  • ✓Conducted comprehensive analysis of 5+ major blockchain platforms including technical architecture comparison
  • ✓Gained hands-on experience with Web3.js and Solidity smart contract development

Technologies Used

Blockchain AnalysisEthereumBitcoinBinance Smart ChainSmart ContractsConsensus AlgorithmsTokenomics

Interested in Collaborating?

Let's connect and discuss potential opportunities or projects