HKUST Fintech Research Project

● Automated text classification, replacing manual processes with NLP models, achieving 90%+ accuracy.

● Fine-tuned GPT-2 and BERT on NVIDIA DGX using HuggingFace, optimizing performance for text classification tasks.

● Led an end-to-end industry report for HKMA, leveraging techniques such as hierarchical clustering from analysis to publication.

● Led a federated machine learning project for a public entity, enhancing model explainability using techniques like SHAP to ensure transparency and interpretability.