people
members of the lab or group
layout: about title: about permalink: / subtitle: Viettel AI.
profile: align: right image: my_avt.jpg image_circular: false # crops the image to make it circular # more_info: > # <p>555 your office number</p> # <p>123 your address street</p> # <p>Your City, State 12345</p>
selected_papers: true # includes a list of papers marked as “selected={true}” social: true # includes social icons at the bottom of the page
announcements: enabled: true # includes a list of news items scrollable: true # adds a vertical scroll bar if there are more than 3 news items limit: 5 # leave blank to include all the news in the _news folder
latest_posts: enabled: false scrollable: true # adds a vertical scroll bar if there are more than 3 new posts items limit: 3 # leave blank to include all the blog posts —
I am an AI researcher and engineer from Vietnam, currently working at the intersection of generative modeling, multimodal AI, and controllable visual generation.
My research journey began with visual text intelligence, especially OCR, handwriting recognition, and mathematical expression recognition. These problems taught me how difficult it is for AI systems to understand fine-grained visual patterns, spatial structure, and content constraints. More recently, my work has shifted toward generative modeling. In my latest research, I study one-shot handwriting generation with diffusion models, focusing on how to capture complex writer styles from a single reference image while preserving textual content and local visual details.
Beyond academic research, I have spent several years building AI systems in industry. At Viettel AI, I have worked on OCR, eKYC, document processing, handwriting-related problems, information extraction, and generative data synthesis for real-world applications. This experience has shaped my research style: I care not only about proposing new models, but also about building systems that are robust, scalable, and useful in practice.
I am now interested in broader questions in compositional and controllable generative AI. How can generative models compose multiple constraints at inference time? How can style, content, structure, and realism be represented modularly? How can diffusion models, energy-based models, and multimodal representations be combined to build more flexible generative systems?
My long-term goal is to contribute to the next generation of generative AI systems: models that are not only high-quality, but also controllable, interpretable, efficient, and useful across real-world domains.
layout: about title: about permalink: / subtitle: Viettel AI.
profile: align: right image: my_avt.jpg image_circular: false # crops the image to make it circular # more_info: > # <p>555 your office number</p> # <p>123 your address street</p> # <p>Your City, State 12345</p>
selected_papers: true # includes a list of papers marked as “selected={true}” social: true # includes social icons at the bottom of the page
announcements: enabled: true # includes a list of news items scrollable: true # adds a vertical scroll bar if there are more than 3 news items limit: 5 # leave blank to include all the news in the _news folder
latest_posts: enabled: false scrollable: true # adds a vertical scroll bar if there are more than 3 new posts items limit: 3 # leave blank to include all the blog posts —
I am an AI researcher and engineer from Vietnam, currently working at the intersection of generative modeling, multimodal AI, and controllable visual generation.
My research journey began with visual text intelligence, especially OCR, handwriting recognition, and mathematical expression recognition. These problems taught me how difficult it is for AI systems to understand fine-grained visual patterns, spatial structure, and content constraints. More recently, my work has shifted toward generative modeling. In my latest research, I study one-shot handwriting generation with diffusion models, focusing on how to capture complex writer styles from a single reference image while preserving textual content and local visual details.
Beyond academic research, I have spent several years building AI systems in industry. At Viettel AI, I have worked on OCR, eKYC, document processing, handwriting-related problems, information extraction, and generative data synthesis for real-world applications. This experience has shaped my research style: I care not only about proposing new models, but also about building systems that are robust, scalable, and useful in practice.
I am now interested in broader questions in compositional and controllable generative AI. How can generative models compose multiple constraints at inference time? How can style, content, structure, and realism be represented modularly? How can diffusion models, energy-based models, and multimodal representations be combined to build more flexible generative systems?
My long-term goal is to contribute to the next generation of generative AI systems: models that are not only high-quality, but also controllable, interpretable, efficient, and useful across real-world domains.