Projects

Things I've built

A selection of research and engineering work spanning graph neural networks, GenAI, sports analytics, and applied ML.

TAGON — Graph Neural Networks for Sequential Recommendation

GNN Temporal Attention PyTorch

Temporal Attention Graph-Optimised Network modelling dynamic user interactions, improving relevance and timeliness by 20% across five real-world datasets and outperforming SOTA baselines in precision, recall, and user satisfaction.

Icarus AI — GenAI Job Recommendation

LLM Explainability Full-Stack

Explainable job recommendation platform using candidate CVs and LLMs to deliver personalised career suggestions. Earned Third Place at the NTU Deep Learning Week Hackathon.

Speech Emotion Recognition with Bi-GRU + Attention

MELD Multi-modal Attention

Bi-directional GRU with attention mechanisms on the MELD dataset, improving classification accuracy beyond single-utterance baselines for multi-party dialogue and cross-utterance emotional dynamics.

Expected Goals Model — Singapore Football

Sports Analytics Ensembles NN

Spatio-temporal xG framework combining KNN, Random Forest, Logistic Regression, and neural networks with ensemble methods. Achieved 99.58% accuracy on a 1M-point English Premier League dataset. NTU President Research Scholar.

Delphi — News Credibility via GCNs

GCN Word2Vec AWS

End-to-end misinformation detection pipeline using GCNs over Twitter propagation patterns, achieving 95% accuracy. Awarded "Innovative Use of Natural Language Processing" distinction.