English 
NVIDIA has released two synthetic datasets to support the development of AI agents tailored to Japan’s language, culture, and industry requirements. Announced at its Agentic AI Day event in Tokyo on November 10, the datasets are Nemotron Personas Finance Japan and Nemotron PII Japan.
The first expands NVIDIA’s existing Japanese persona dataset to 9 million synthetic profiles, adding financial attributes such as income, assets, spending, savings, and investment behavior. The profiles can be used by financial companies for simulations on their clients, market research, and testing new AI agents.
こちらもお読みください： Smartsheet、AIエージェント、連携機能、およびセキュリティ更新を発表
A second set of 100,000 synthetic Japanese documents with eight sectors and 43 types of document has been created for training and evaluation of sensitive information detection and masking models.
The datasets are publicly accessible on Hugging Face. The release tackles an increasing enterprise AI challenge making industry-specific training data available to organizations without revealing real customers’ data.
日本語