On October 23, at the working program of the high-ranking Vietnamese delegation to the Republic of Bulgaria to expand strategic cooperation space between the two countries, Mr. Huynh Quang Liem, General Director of VNPT Group, shared about the strategy of mastering core technologies. Specifically, VNPT is strongly transforming from a traditional telecommunications service provider to a digital technology enterprise, in which AI Make in Vietnam based on Vietnamese data is considered the focus of investment to affirm Vietnams AI autonomy and sovereignty.
Mr. Liem also cited VNPT's milestones in science and technology development in the field of AI, typical in recent times with research projects published at world-leading AI conferences such as MICCAI, EMNLP, ICASSP, ... This is one of the platforms for VNPT to develop the "Make in Vietnam" ecosystem to serve tens of millions of Vietnamese people, with a scale of processing billions of user requests.
Specialized AI for Vietnam starts with Vietnamese data
In medicine, incorrect or slow diagnosis can affect the quality of treatment, costs and psychology of patients. Diseases such as thyroid cancer often progress silently, while the number of specialists and diagnostic equipment in Vietnam, especially lower-level medical facilities, is still limited. From that practice, the VNPT group of engineers coordinated with doctors at Hanoi Central Endocrinology Hospital to conduct research and develop AI to support the diagnosis of thyroid cancer from ultrasound images, specializing in Vietnam. Recently, the research results were announced at the MICCAI 2025 conference, taking place from September 23 to 27, 2025 in Korea. This is the leading international conference on AI and computer vision in medicine, focusing on the use of AI technology to analyze medical images.
Specifically, the study focused on the application of AI in the diagnosis of thyroid cancer, conducted with data from nearly 10,000 patients in 3 regions of the country over 4 years. The dat sets are labeled by leading experts and doctors themselves, authenticated by pathology tissue and standardized formats to ensure clinical reliability.
This study plays a fundamental role in the development of automatic diagnostic support systems suitable for the characteristics of the population and medical conditions in the country, helping to improve accuracy, shorten diagnosis time, reduce the burden on doctors, and expand access to high-quality medical services to facilities. With a large scale and high standardization, the project is an important source of training and assessment data for AI models in medical imaging diagnosis, helping to improve the accuracy of detecting and classifying thyroid damage in Vietnam.
Bringing Vietnamese to the AI map of global language processing
While popular languages such as English or Chinese possess rich sources of data, Vietnamese is still among the language groups with limited resources, making the development of language AI models more challenging. Due to the lack of standardized data, many chatbot systems or virtual assistants today still give the experience of not understanding the context correctly, expressing awkwardly and unnaturally when communicating in Vietnamese. Faced with that reality, VNPT's AI engineers have focused on researching and developing Vietnamese language and emotions models. These studies were also announced by the team at the EMNLP 2025 conference taking place from November 4-9, 2025 in China - the world's A*-class AI conference on natural language processing (NLP).
These projects apply advanced machine learning techniques in developing large language models (LLM) and generative AI (Generative AI), aiming to improve the ability to understand the context, emotions and Vietnamese nuances of AI models - which are complex factors and are less modeled in international language sets. Thanks to that, the Vietnamese language is increasingly present in the global NLP research community, opening up the ability to build strong, identity-rich language models that are suitable for Vietnamese people.
In parallel, at ICASSP 2025, taking place from April 6 to 11, 2025 in India - the conference on AI voice processing ranked A1, VNPT's research team introduced a research project on voice conversion technology, combined with the regional competitive learning method (domain adversarial training) - a technique that helps the learning model to stably retain Vietnamese voice characteristics even when training data is limited or comes from many different sources. Thanks to that, the system can create a natural voice, closer to humans, while flexibly customizing to each user to bring a more friendly and personalized communication experience.
When applied in practice, this technology will overcome the major limitations that many systems are currently facing, which are monotonous, emotionless AI voices and are " machine-based", helping users feel closer and more confident when interacting with AI.
From scientific knowledge to national technological capacity
International research works are an important intermediary step in VNPT's AI value chain, where scientific knowledge is transformed into practical products. The above models and dataseters are currently applied in many fields such as Vietnamese voice recognition, computer vision for transportation, healthcare, public administration virtual assistants, and recently AI Agent for banking and finance.
These AI Agent are built on the platform of Vietnamese language models and technology to understand dialogue context, helping to automate customer support processes, financial product consulting, transaction authentication support, while ensuring high requirements for data security and compliance - a particularly important factor in the finance - banking sector.
Research works published at international conferences of the Vietnamese scientific team, including VNPT, show that domestic research capacity is gradually approaching global standards. More importantly, this is the first step for the country's greater goal: forming the ability to master sovereign AI technology, based on data, language and knowledge of the Vietnamese people themselves.
●MICCAI ( International Conference on Medical Image Computing and Computer-Assisted Intervention) - a leading prestigious international conference on AI and computer vision in medicine, focusing on medical imaging calculation and computer-assisted intervention. According to the CORE 2023 rating, MICCAI is ranked A in the field of Applied Computing and computer vision / multimedia.
●ICASSP (IEEE International Conference on Acoustics, Speech, and Signal Processing) - a prestigious international conference, ranked A1, on signal, sound and voice processing, organized annually by IEEE with a history of more than 50 years.
● EMNLP (Conference on Empirical Methods in Natural Language Processing) - a leading prestigious international conference on natural language processing (NLP), stands out with experiential approaches and strict acceptance rates. In the field of NLP, EMNLP is a top conference, ranked A*.
