Education
University of Oregon
Ph.D. in Computer Science, GPA 4.16/4.0, Excellent Degree. Research Topic: Efficient LLMs. Co-advised by Assoc. Prof. Thien Huu Nguyen and Asst. Prof. Jieyang Chen.
Ph.D. in Computer Science, GPA 4.16/4.0, Excellent Degree. Research Topic: Efficient LLMs. Co-advised by Assoc. Prof. Thien Huu Nguyen and Asst. Prof. Jieyang Chen.
Fall 2024 - Present
Hanoi University of Science and Technology
B.S. in Computer Science, GPA 3.71/4.0, Excellent Degree. Advised by Dr. Linh Ngo Van.
B.S. in Computer Science, GPA 3.71/4.0, Excellent Degree. Advised by Dr. Linh Ngo Van.
Fall 2018 - Spring 2023
Employment
Microsoft Research
Research Intern. Engineered and optimized "Megakernels" on NVIDIA GPUs to accelerate LLM inference and reduce memory-bound latency. Designed low-level scheduling methods and custom fused operators using CUDA/Triton, maximizing hardware utilization.
Research Intern. Engineered and optimized "Megakernels" on NVIDIA GPUs to accelerate LLM inference and reduce memory-bound latency. Designed low-level scheduling methods and custom fused operators using CUDA/Triton, maximizing hardware utilization.
Jun 2025 - Sep 2025
Redmond, WA, USA
Redmond, WA, USA
VinAI Research
Research AI Resident on the Natural Language Processing Group. Worked on Information Extraction tasks and Large Language Models (LLMs).
Research AI Resident on the Natural Language Processing Group. Worked on Information Extraction tasks and Large Language Models (LLMs).
Jul 2022 - Jul 2024
Hanoi, Vietnam
Hanoi, Vietnam
Publications
CRISP: Cliff-awaRe Input-adaptive Sparse Prefilling with Structural-Mass-Motivated Routing
Input-adaptive sparse attention for long-context LLM prefilling, with up to 5.30x attention speedup at 512k tokens.
EMNLP 2026
Lizard: An Efficient Linearization Framework for Large Language Models
An efficient linearization framework for LLMs.
ACL 2026
Towards Fast and Accurate Modeling for Cross-Lingual Label Projection
Fast and accurate modeling for cross-lingual label projection.
ACL 2026
Massively Multilingual Instruction-Following Information Extraction
Multilingual instruction-following for information extraction.
ACL 2025 Findings
Taipan: Efficient and Expressive State Space Language Models with Selective Attention
Efficient and expressive state space language models with selective attention.
Arxiv
Transitioning Representations between Languages for Cross-lingual Event Detection via Langevin Dynamics
We explored a novel alignment method for cross-lingual transfer learning in Event Detection.
EMNLP 2023 (Findings)
A Spectral Viewpoint on Continual Relation Extraction
A novel method for Continual Relation Extraction (CRE) with Feature Decorrelation.
EMNLP 2023 (Findings)