YTL AI Labs Launches 1.35 Million AI Personas

YTL AI Labs has launched Nemotron-Personas-Malaysia, an open dataset containing 1.35 million synthetic personas designed to help developers build and test artificial intelligence systems for Malaysian users.

 

The dataset was developed with NVIDIA and generated from 150,000 base records using probability distributions grounded in official Malaysian demographic statistics.

Unlike a database of real people, the personas are artificially generated profiles. YTL AI Labs said the dataset contains no personal data and does not identify any individual.

Nemotron-Personas-Malaysia is available for commercial use under the CC BY 4.0 licence. It is intended for developers, researchers and organisations that need representative Malaysian user profiles for AI development, testing and evaluation.

The dataset covers 39 fields, including age, gender, location and occupation. It also incorporates personality characteristics using the five-factor OCEAN model, which covers openness, conscientiousness, extraversion, agreeableness and neuroticism.

Its demographic distributions are drawn from published information from the Department of Statistics Malaysia, OpenDOSM, MyCensus, eStatistik and national labour force publications.

YTL AI Labs said the dataset preserves demographic differences at regional and district level instead of treating Malaysia as a single uniform population.

This includes Malay, Chinese and Indian populations as well as groups such as Kadazan-Dusun, Bajau, Murut, Iban, Bidayuh and Melanau. Regional Bumiputera categories across Sabah, Sarawak and Peninsular Malaysia are also represented.

The approach is designed to give AI developers synthetic users with combinations of characteristics that more closely follow Malaysian population patterns.

This could be useful when testing whether an AI application behaves consistently across different locations, occupations, age groups, languages and socioeconomic backgrounds.

For example, developers could create synthetic customer interactions involving people from different Malaysian states and occupations instead of repeatedly testing an AI service against generic user profiles.

Financial institutions could use similar profiles when evaluating how an AI system interprets different ways customers describe financial requirements. Government agencies could also test digital services across synthetic users representing different demographic groups.

The personas can also be used for synthetic data generation, fine-tuning, alignment work, red-teaming and application testing.

NVIDIA describes its wider Nemotron-Personas project as a collection of multilingual, region-specific synthetic persona datasets grounded in real demographic and geographic distributions. Its NeMo Data Designer framework supports statistical distributions, correlations between fields and reproducible synthetic data generation.

Nemotron-Personas-Malaysia uses a staged dependency structure in which certain persona characteristics influence other attributes. This is intended to produce profiles where demographic, employment, regional and personality characteristics remain internally consistent rather than being generated independently at random.

The dataset also uses NVIDIA’s NeMo Data Designer technology and the Nemotron-Personas framework. Developers are not restricted to the attributes included in the Malaysian release and can use the same methodology to create synthetic personas tailored to particular applications.

Nemotron-Personas-Malaysia joins NVIDIA datasets developed for markets including the United States, Japan, India, Singapore, Brazil, France, Korea, El Salvador, Vietnam and Belgium.

YTL AI Labs said the Malaysian release is the first dataset in the Nemotron-Personas collection to lead with Bahasa Melayu.

“The next generation of AI will not be defined only by who builds the largest model, it will be defined by how well intelligence understands the people it serves. Malaysia has its own languages, cultures, institutions and ways of working. If AI is going to become part of our everyday lives, we need to make sure that Malaysian context is represented from the ground up,” YTL AI Labs CEO Foong Chee Mun said.

“Nemotron-Personas-Malaysia gives our entire ecosystem an open foundation to build on,” he added.

The release also expands YTL AI Labs’ work with NVIDIA. In March 2026, the companies announced ILMU-Nemo-30B, an agentic AI model fine-tuned using Malaysian datasets and NVIDIA Nemotron technology. The model was trained using infrastructure hosted on YTL AI Cloud.

YTL AI Labs is developing the ILMU family of Malaysian AI models and services, while YTL AI Cloud provides computing infrastructure for training and operating AI systems in Malaysia. YTL describes ILMU as being designed around Malaysian languages, cultural context and local use cases.

Nemotron-Personas-Malaysia adds a population dataset to that stack, giving developers synthetic Malaysian profiles that can be used when developing or evaluating locally focused AI systems.

The dataset is available openly through Hugging Face and is compatible with NVIDIA NeMo tools.