← AI news

Nvidia released a Japanese synthetic dataset for finance AI

Nvidia released a Japanese synthetic dataset for finance AI
Picture from Hugging Face
9M
Personas in the dataset
50
Fields per record
1600+
Occupation categories
Who it's for
AI developers in Japan
Where it works
Hugging Face
When
Released November 2026
Open Hugging Face

Key points

  • 9 million personas from official statistics
  • No real customer data used
  • Written in natural Japanese
  • Open license, commercial use allowed
  • Works with NeMo User Sim

Where to use it

  • Customer-service AITest agents on simulated customer conversations
  • Conversation designUse personas as seeds for dialogue
  • Fairness testingCheck AI performance across regions and ages

What changed

  • Fields per record2250

What it does

A large set of synthetic Japanese personas with household finances and financial literacy details, for building and testing AI.

Who it's for

Banks, insurers and developers building AI for Japanese users.

Price and where to try it

The item gives no price. It is released under the CC BY 4.0 license, which allows commercial use, and is on Hugging Face.

Source

Written with AI from the source above. If something looks off, the source is what counts.

Read next