yte ai labs and nvidia launch 1.35 million synthetic Malaysian personas dataset

KUALA LUMPUR, 23 SEPTEMBER 2026 —

YTL AI Labs and Nvidia have released Nemotron‑Personas‑Malaysia, a dataset of 1.35 million synthetic Malaysian personas that developers can use to train and test artificial‑intelligence applications for local users, the companies said in a joint statement on Wednesday.

The dataset was built from 150,000 base records and expanded using probability distributions drawn from published Malaysian demographic statistics, YTL AI Labs explained. It contains 39 fields covering age, gender, occupation, location and personality traits, and is hosted on the Hugging Face platform under a Creative Commons Attribution 4.0 licence that permits commercial use with attribution.

YTL AI Labs CEO Foong Chee Mun stressed that “Malaysian languages, cultures and ways of working need to be represented in the data used to develop AI applications,” a comment made in the company’s press release. The personas are designed to mirror demographic patterns down to the district level, reflecting ethnic and regional differences across Peninsular Malaysia, Sabah and Sarawak, according to the statement.

Because the synthetic records contain no personal data, they cannot be used to identify any individual, YTL AI Labs added. The firm said developers can employ the personas to generate synthetic training data and to test whether AI systems behave differently for various user groups. Potential use cases cited include customer‑service chatbots, banking interfaces and government digital services, all of which rely on accurate representation of Malaysia’s diverse population.

Nemotron‑Personas‑Malaysia is compatible with Nvidia’s NeMo libraries and marks the first entry in Nvidia’s Nemotron‑Personas collection to lead with Bahasa Melayu, YTL AI Labs noted. The collaboration leverages Nvidia’s expertise in large‑scale language models and YTL AI Labs’ local market knowledge, creating a resource that bridges global AI technology with Malaysian cultural nuances.

While the release does not directly involve property transactions, the dataset could influence the real‑estate sector’s digital transformation. Property portals and developers that use AI‑driven recommendation engines or virtual assistants may adopt the personas to ensure their tools respond appropriately to users from different states, income brackets and language backgrounds. By testing algorithms against a representative synthetic population, firms can reduce bias and improve the relevance of property suggestions, potentially boosting conversion rates.

Analysts at EdgeProp highlighted that the move reflects a broader trend of Malaysian tech firms seeking to localise AI solutions. With the government’s push for digital adoption across industries, tools such as Nemotron‑Personas‑Malaysia provide a practical way for fintech, proptech and e‑government platforms to comply with data‑privacy regulations while still accessing rich, demographically accurate training data.

The dataset is available immediately on Hugging Face, and YTL AI Labs encourages developers to attribute the source when using it commercially. As more Malaysian companies integrate AI into their workflows, the synthetic persona library may become a standard benchmark for evaluating fairness and performance across the nation’s multi‑ethnic market.

Related: YTL AI Labs · Foong Chee Mun · Kuala Lumpur