I am a PhD candidate at the Applied Artificial Intelligence Institute (A²I²), Deakin University, supervised by Prof. Truyen Tran and Dr. Hung Le. My research focuses on Sparse Mixture-of-Experts (MoE) for the efficient training and inference of large language models, and on controllable and robust reasoning in LLMs. A unifying thread of my PhD is the binding problem: how foundation models compose, separate, and route information across modular components.
Before starting my PhD, I studied an M.Sc. in Computer Science at the University of Tennessee at Chattanooga (UTC) and an M.Sc. in Data Science at VNU University of Science. I have also spent more than six years as a data scientist and AI researcher in industry, including roles at Panasonic R&D Vietnam, Techcombank, and Toyota Motor Vietnam.
Education
PhD in Information Technology, 2024–present, Deakin University, Australia
M.Sc. in Computer Science, 2023–2024, The University of Tennessee at Chattanooga (GPA 4.0/4.0, class rank 1st)
M.Sc. in Data Science, 2020–2022, VNU University of Science (GPA 3.75/4.0)
B.Tech. in Information Technology, 2019–2021, Hanoi University of Science & Technology (top 7%)
B.A. in International Business Economics, 2010–2014, Foreign Trade University
Selected Recent Papers
- Eigenvectors of Experts are Training-free Non-collapsing Routers, ICML 2026 (Spotlight).
- Rethinking Sparse Mixture of Experts from a Unified Perspective, ICML 2026.
- Do Domain-specific Experts Exist in MoE-based LLMs?, ACL 2026 (Findings).
- SimSMoE: Toward Efficient Training of Mixture of Experts via Solving Representational Collapse, NAACL 2025 (Findings).
- On the Role of Discrete Representation in Sparse Mixture of Experts, TMLR 2025.
- HyperRouter: Towards Efficient Training and Inference of Sparse Mixture of Experts, EMNLP 2023.
See the Research page for the full list.
Resources
A CV (PDF) and code/projects on GitHub.
Contact
Email: giangdo.utc [at] gmail.com
Office: Applied AI Institute (A²I²), Deakin University, Geelong, VIC, Australia