Nvidia and FPT release 900K synthetic personas dataset for Vietnam

Nvidia and FPT Corporation have released a dataset of 900,000 synthetic personas designed to help AI models understand Vietnam’s language, culture, and demographics. The Nemotron-Personas-Vietnam dataset, launched on June 5, dropped on Hugging Face under a CC-BY-4.0 license, meaning it’s commercially usable by anyone.

What’s actually in the dataset

The collection spans 31 fields per persona, covering Vietnamese demographics, geographic distribution, language diversity, and labor characteristics. These aren’t scraped profiles from real individuals. They’re algorithmically generated to reflect genuine population patterns while sidestepping the privacy minefield that comes with using real personal data.

The dataset is compatible with Nvidia’s NeMo tools, the company’s framework for building and customizing AI models. FPT Corporation, which operates as an Nvidia Cloud Partner, brought the local expertise needed to make the personas culturally and linguistically accurate.

The sovereign AI play

This release is part of Nvidia’s broader Nemotron-Personas initiative, which has already produced similar region-specific datasets for Singapore, Korea, and the US. The launch coincided with Nvidia GTC Taipei and Computex 2026, two of the biggest events on the Asian tech calendar.

Nvidia’s partnerships extend beyond FPT in the country. Viettel, another major Vietnamese tech firm, is involved in building national AI applications on Nvidia’s infrastructure. FPT’s role as an Nvidia Preferred Partner also extends beyond Vietnam, with the company enhancing AI factories in both Vietnam and Japan.

What this means for the AI and tech landscape

By making the dataset freely available for commercial use under CC-BY-4.0, Nvidia and FPT are providing startups, universities, and smaller companies with 900,000 personas to work with at no cost. Synthetic data generation also sidesteps increasingly strict data protection regulations, offering a compliance-friendly alternative to using real personal data in AI training.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.