FPT and NVIDIA Expand Nemotron-Personas-Vietnam Dataset by 10X Times to Power the Next Wave of Agentic AI

HANOI, Vietnam, Oct. 16, 2026 /PRNewswire/ -- FPT announced the expansion of Nemotron-Personas-Vietnam, its open dataset of synthetic personas grounded in Vietnam's official demographic and labor statistics. The dataset now contains 1,000,000 personas, giving developers, researchers, and enterprises a larger and more representative foundation for building models, agents, and applications that understand Vietnamese people, language, and culture.

At a Glance

First released by FPT in collaboration with NVIDIA in June 2026, Nemotron-Personas-Vietnam is an open dataset of synthetic personas grounded in Vietnam's official statistics from the General Statistics Office (GSO), including the Population and Housing Census 2024 and the Vietnam Household Living Standards Survey (VHLSS) 2024, as well as from FPT Smart Cloud and Quantum AI & Cyber Security Institute (FPT Corporation).

• 21 fields: 6 persona fields and 15 contextual fields covering persona attributes and demographics.
• 13,000 unique full-name combinations.
• 20 occupational categories derived from labor survey data.

Built on NVIDIA's Nemotron-Personas methodology, which anchors persona generation in real-world statistics and uses open-weight LLMs to write high-fidelity persona narratives, the dataset gives developers realistic Vietnamese user profiles spanning age, region, occupation, income, education, and cultural context, all without any personally identifiable information (PII).

A Dataset That Understands Vietnam at Scale

This expansion is more than a larger dataset. Growing to 1,000,000 personas opens new ways for developers to build, train, and evaluate AI for Vietnam:

• Finer segmentation: Users can sample highly specific slices of the Vietnamese population, such as personas that represent a retiree in Ho Chi Minh City, a smallholder farmer in the Mekong Delta, or a young engineer in Da Nang, with enough records in each to train and test reliably.
• Larger workloads: The expanded scale supports bigger training runs, more comprehensive red-teaming, and population-level simulations that the original release could not sustain on its own.
• Reduced model collapse: Richer variation in narratives, skills, and behaviors lowers the risk that models trained on the data collapse onto a narrow set of templates, a growing concern as synthetic data becomes load-bearing in AI pipelines.
• Deeper coverage: the earlier release sampled a set of major provinces and cities. The 1,000,000-persona version covers the whole country, so developers can build and test for users in every region.

The dataset remains fully synthetic, contains zero personally identifiable information, and is released openly under a permissive license on Hugging Face, compatible with NVIDIA NeMo libraries.

A Commitment to Vietnam's Sovereign AI

An expansion sends a different signal than a launch. It tells developers that the resource they started building on is being maintained, grown, and invested in, not left as a one-time release.

For FPT, this is a deliberate statement of long-term commitment. Vietnam is accelerating its digital transformation. Under the newly approved Program for Digital Economy and Digital Society Development 2026-2030, the government targets the digital economy to reach approximately 30% of GDP by 2030. Across banking, finance, healthcare, insurance, retail, and public services, organizations increasingly need realistic persona data for product testing, service personalization, simulation, and AI model evaluation. By growing the dataset rather than leaving it as-is, FPT is building the data layer of Vietnam's sovereign AI, AI grounded in the country's own population, language, and values rather than imported from elsewhere.

Fostering Agentic AI Capabilities in Practice

As agentic AI emerges as the next frontier, systems that reason, act, and serve specific users rather than just answer prompts, the demand for grounded, identity-aware data will only grow. A larger, more representative persona dataset gives the Vietnamese AI community the foundation to build agents that genuinely understand who they are serving.  

Customer Personas: Build customer-facing agents that can draw on personas spanning income brackets, regions, and financial-literacy levels to adjust their  tone, product recommendations, and compliance handling for each customer, such as a retiree in Hanoi versus a young professional in Ho Chi Minh City.

Simulation and Evaluation for Agentic Systems: Create simulation environments populated with realistic Vietnamese users and organizations, run red-teaming and bias-and-fairness evaluations across diverse demographic groups, and stress-test how agents behave across regions, ages, and occupations, all without touching real customer data.

Domain-Specific AI Agents: Enable specialized AI agents across regulated industries, including finance, healthcare, insurance, and public services, that need to reason about who they are helping. A healthcare agent can be evaluated against patient personas spanning literacy levels, regions, and care settings. A public-service agent can be tested against citizen-personas across the country's administrative landscape.

Download the dataset for free on Hugging Face.