curriculum vitae

Ph.D. in Computer Science (ML/AI)

• Advised by Prof. Anh Nguyen
• GPA: 3.91/4.0
• Thesis: Improving Interpretability and Accuracy of Artificial Intelligence in Natural Language and Image Understanding

B.S. (Honours) in Computer Science and Engineering

• Advised by Dr. Son Tran
• GPA: 3.39/4.0
• Thesis: Scene Text Detection and Recognition (Distinction)

Research Scientist Intern at Adobe Reseach

• Quantizes and deploys LLMs (∼0.1–7 billion parameters) locally on mobile devices for benchmarking.
• Proposes SlimLM models (∼0.05–3 billion parameters), pre-trained from scratch on ∼1–5 trillion tokens using datasets such as SlimPajama, FineWeb, DCLM or SmolLM.
• Finetunes and evaluates SlimLM and baseline models for Q/A, question suggestion and summarization for documents.
• Develops a mobile application for PDF reader with AI assistant to demonstrate the efficiency of SlimLM on documents.

Research Scientist Intern at Adobe Reseach

• Crawled and analyzed trends in Reddit's image editing requests.
• Constructed an instruction-following dataset and a human preference data to teach large language models (LLMs) to use visual foundation models for handling image editing requests.
• Fine-tuned open-source LLMs (e.g., LLaMA-2, Mistral, Zephyr, Falcon, MPT) across different versions (7B, 13B, 70B) to correctly generate and execute a plan including a sequence of image editing tools to handle simple, complex and implicit user requests for Adobe products (e.g., Photoshop).

Research Scientist Intern at Adobe Reseach

• Developed a deep neural model for phrase representation.
• Designed a human annotation guideline and led the team (∼15 people) to construct a dataset and benchmark to enhance and evaluate in-context phrase representation for semantic search.
• Fine-tuned and evaluate small and large language models in understanding phrases in context.
• Developed a smart phrase search system that is potentially utilized in the Acrobat product.

ML/AI Research Engineer at AIST Japan

• Investigated state-of-the-art deep learning models for named entity recognition (NER), relation and event extraction in biomedical domain.
• Implemented novel deep neural networks for NER task with Dr. Sohrab (EMNLP 2018) and Dr. Ju (NAACL 2018).
• Developed the DeepEventMine system with Dr. Trieu in the first phase of the project.
• Fine-tuned hyper-parameters with greedy search and Bayesian optimization methods.
• Pre-processed biomedical corpora (JNLPBA2004, GENIA, ACE2005, CG2013, MLEE, PHAEDRA).
• Optimized parallel computing with multiple GPUs for speeding up a model training process.
• Collaborated with Dr. Nagano to create the EzCat database.
• Developed pipelines to evaluate EventMine systems for NaCTeM (University of Manchester).

Senior Software Engineer at OPSWAT Vietnam (now Beowulf)

• Developed over-the-top mobile applications (VoxyPAD, Tutorica, Victoria and Hana).
• Worked directly with CEO Dr. William Nguyen and other team leaders to design APIs for the whole systems.
• Supported clients in the United States and in Japan (Toshiba corporation) to deploy our products.

Software Engineer at VoxyPAD Vietnam (now Beowulf)

• Developed a back-end system using Spring framework to provide user account management for virtual RADIUS servers.
• Developed a socket server for an over-the-top application to send and receive messages via socket.

• Diversity and Inclusion Award, EACL 2023.
• Graduate Student Council Travel Fellowship.

• Auburn University Graduate Assistantship.

• University of Science Faculty of IT Excellence Fellowship.

• Top-3 university entrance award from high school.

• I am a reviewer for the following conferences: NeurIPS, ACL, EMNLP, NAACL.

• I was a reviewer for the following conferences: NeurIPS, ICLR, ACL, EMNLP, EACL.

• I was a youth program personnel for K-6 AI Club.

• I was a reviewer for the following conferences: NeurIPS, EMNLP.