NVIDIA Open-Sources Nine Million Synthetic Japanese Finance Personas

By Daniel Okafor

Elith and NVIDIA released Nemotron-Personas-Finance-Japan on November 9, 2026, opening nine million synthetic Japanese personas for commercial and non-commercial AI development under a CC BY 4.0 license. Each persona pairs demographic data with household finances, money behavior and financial-literacy scores, giving banks, insurers and developers a way to test customer-facing systems without touching real customer records.

NVIDIA describes the release, to its knowledge, as the first open persona dataset at million-record scale to combine Japan's official statistics with household-finance and financial-literacy data. The numbers behind that claim are concrete: one million Japanese-language records, nine personas per record, 50 fields per record, coverage of all 47 prefectures and more than 1,600 occupation categories.

What does the dataset contain?

The release extends Nemotron-Personas-Japan, whose September 23, 2025 version held six million personas across 22 fields and excluded finance-specific personas. The new edition adds real financial texture:

• Three of the 20 persona-related fields describe financial habits, decision-making behavior and financial concerns separately
• The 30 contextual fields cover personal income, household assets, housing and mortgage status, consumption propensity, financial-literacy scores, internet and smartphone use, household structure and self-rated health
• Household-asset distributions draw on the Statistics Bureau's 2019 National Survey of Family Income, Consumption and Wealth

That structure supports multi-turn Japanese dialogue simulation. NVIDIA's examples run from a regional pensioner asking about asset succession to an urban dual-income household reviewing education savings and a mortgage.

The age-related asset data matters for product design. The dataset card's analysis reports that the share of households with at least 20 million yen in financial assets rises from about 10% at ages 25–29 to about 30% at ages 60–64. Assets stay relatively high through people's 60s and 70s even as personal income falls around retirement. That pattern supports distinct customer profiles for wealth building, retirement planning and inheritance.

Why Japan's demographics shape the use cases

The demographic context is especially relevant for retirement and inheritance simulations. The Statistics Bureau's 2024 population estimates put people aged 65 or older at 29.3% of Japan's population as of October 1, 2024. The synthetic personas cover only adults aged 18 and older.

Intended uses include customer-service agents, sales and marketing conversation tests, NeMo User Sim simulations, red-teaming and checks for consistent behavior across rural and urban locations, ages, education levels and economic circumstances.

How were the personas built?

The generation stack combines three components:

• NeMo Data Designer, which supplies structured outputs, validation and automated retries
• The NeMo SDG-PGMs probabilistic-graphical-model framework
• The Apache-2.0-licensed gemma-4-31B-it model for Japanese text

Sompo Japan, Tokio Marine Holdings, Bank of Fukuoka, Mizuho Financial Group, Meiji Yasuda Life Insurance and the Financial Data Utilization Association provided feedback during the design process. The dataset card also names J-FLEC's survey and statistical data as a seed input. J-FLEC's official index says its financial-literacy survey covers people aged 18–79, while its household financial-behavior surveys examine assets, liabilities and life planning for two-person and single-person households.

What are the privacy safeguards?

Every persona is artificial rather than a record of a real customer or individual, and any resemblance to an actual person is coincidental. The public Hugging Face version omits standalone first- and last-name fields and synthetic addresses, and excludes healthcare personas. The card discloses an assumption that occupation is independent of education given prefecture and sex, and notes that comprehensive census statistics on gender independent of sex are unavailable.

What is the companion PII dataset?

NVIDIA paired the release with Nemotron-PII-Japan, a companion dataset for training and testing systems that detect and redact personally identifiable information. It contains 100,000 synthetic Japanese business documents spanning 43 document types in eight domains, with 56 PII labels and 810,703 annotated spans. The finance personas can simulate customers while the PII documents test whether a pipeline identifies sensitive data.

The PII card carries a caution worth reading closely. Good results on synthetic documents, it says, do not guarantee complete detection, fairness, privacy protection, regulatory compliance or prevention of leakage on real data.

The two releases give financial AI teams an open way to test both customer interactions and privacy controls. Real-world performance against institutions' own Japanese-language workflows remains the decisive measure of their practical value.