NVIDIA opens Nemotron Personas Finance Japan synthetic data

NVIDIA positions these datasets as part of Nemotron, its system for building and deploying AI agents at lower cost while balancing accuracy and compute efficiency. The release addresses a practical obstacle: AI agents need to reflect a country's language, culture, demographics and industry conditions, but personal data is hard to obtain, use and share because of privacy and security concerns, especially in finance, healthcare and the public sector.

Nemotron Personas Finance Japan builds on Nemotron Personas Japan, keeping its basic attributes and personal context while adding finance-specific traits such as income and assets, so that spending, saving, investing and financial decisions can be represented more realistically and diversely. NVIDIA says the dataset draws on public demographic and labor statistics released in 2025 and was created with cooperation from major Japanese financial institutions.

Nemotron PII Japan complements that by supplying documents with annotations showing where personal identifiers sit and which category they belong to. Because documents and correct labels are provided together without any real personal information, teams can fine-tune, evaluate and improve detection and masking models, which may make it easier to build Japanese AI agents that meet local and industry requirements.