Portfolio

Graph Neural Network

Classrooms allocation with Deep Learning

Neural formular structure : An GNN system which for the target relationship.

This AI-powered system optimizes student classroom assignments by modeling social and academic dynamics through Graph Neural Networks (GNN) and Reinforcement Learning (RL).

By treating students as nodes in a social graph, the system captures complex relationships—such as friendships and conflicts—to create balanced learning environments that traditional clustering methods often miss.

Built with a FastAPI and React stack, it features an interactive dashboard for visualizing student networks and allows for real-time manual adjustments, ultimately automating the creation of classrooms that are both academically diverse and socially supportive.

Dependency Parsing: A group allocation system base on GNN

Detect Ai by security auditor, LLM models

First I build This project investigates how psychological manipulation can be used to trigger prompt injection attacks in AI systems, filling a critical gap where no standardized risk framework currently exists.

By utilizing an Agile and DevSecOpsmethodology, the team is developing a structured Risk Map and a Jupyter Notebook-based prototype to visually represent and quantify these vulnerabilities.

The implementation includes an automated test suite with over seven test cases for models

Gemma 3 and Llama 3.1, alongside a JSON database of mitigation strategies to help organizations secure their AI workflows.

Driven by a team of six Data Science students and mentored by experts from CSIRO Data61, this project competition focuses on delivering actionable security insights and a fully functional risk assessment dashboard

Data Engineering

IOT web app

FROM start to end, I have integrate IOT technologies to detect troubleshooting purpose by centralizing big data integreate with the modeling deloy through platform. Finally, we maintain the stucture of big data and improving all business

Data Science

Face Mesh Web App

An interactive web-based application that performs real-time facial landmark detection and hand tracking directly in the browser.

The system uses computer vision and machine learning to map detailed facial features and hand movements, enabling gesture-based interaction and intelligent visual analysis.

Predict Athlete Performance by Ultralytics (Golf Analysis)

In this project I am going to perform a data-driven using techniques inspired by Ultralytics to analyze and predict golf performance.

It leverages metrics like swing speed, launch angle, and shot consistency to model outcomes (e.g., accuracy and distance) and delivers insights through a performance dashboard for targeted improvement.

Contented by me

Besides Data Science, I also have a great passion for content creation. Below is a list of technology and lystlye content, documented to retain beautiful passion intregrate to my lifestlye and amazing people, environment, interest I met on the way.

In Australia, Melbourne - Study with me