cv

PhD Student/Research Assistant

A2I2, Deakin University

• Hallucinations
• Reliability
• Large Language Models
• Agentic LLMs

Data Science Mentor (part-time)

CoderSchool (IT W21)

Data Scientist

XPON Technologies Group (ASX:XPN), Representative Office

My work focuses on developing the data science portion of Wondaris (a Customer Data Platform product)

• Identity resolution
• End-to-end machine learning
• Automate feature suggestion

Senior Data Scientist

Vingroup JSC

My work focused on building data warehouses and solving data science problems of VinHomes (a subsidiary of Vingroup)

• ETL data and building data warehouse
• Analytical reporting
• Entity matching development
• Lead scoring

Data Science Instructor (part-time)

Developer Circles Vietnam

Data Scientist

Viettel Digital

My work focused on building Machine Learning models to improve the operation of products on ViettelPay ecosystem

• Lead generation
• Credit scoring

Machine Learning Engineer

3S Intersoft JSC

My work focused on applying machine learning to solve natural language processing and computer vision problems

• Sentiment analysis
• Object detection, object recognition

Telecommunications Engineer

Viettel Aerospace Institute (VTX)

Research on preliminary design of satellite telecommunications system

Bachelor of Science

Hanoi University of Science and Technology (HUST)

Electronics and Communications Engineering

• Honors Program
• CPA: 3.28 (3rd/22)

First prize in Kalapa Credit Scoring Challenge for Students

Kalapa JSC

Second prize in Vietnam Mathematics Olympiad for University Students

Vietnam Mathematical Society

First prize in Vietnam Mathematics Olympiad for University Students

Vietnam Mathematical Society

Probabilities Are All You Need: A PRobability-Only Approach to Uncertainty Estimation in Large Language Models

AAAI

Uncertainty estimation is vital for mitigating hallucinations in Large Language Models (LLMs), but existing methods such as predictive or semantic entropy remain computationally expensive. We propose PRO, an efficient, training-free uncertainty estimation method that approximates predictive entropy using only the responses' top-K probabilities.

Topic classification of vietnamese product reviews in e-commerce using PhoBERT

Journal of Marketing Analytics

This study uses a semi-supervised NLP approach, combining unsupervised topic modeling with a PhoBERT-based classifier to accurately categorize Vietnamese e-commerce customer reviews.

Representation Learning for Early Sepsis Prediction

Computing in Cardiology

As part of the PhysioNet/Computing in Cardiology Challenge 2019, we propose a neural network called AEC-Net to early detect sepsis based on physiological data.