NVIDIA Launches Nemotron-Personas-Korea Dataset for AI Agent Development

NVIDIA has introduced Nemotron-Personas-Korea, a dataset of 6 million synthetic personas based on official Korean statistics from sources like the Korean Statistical Information Service (KOSIS) and the Supreme Court of Korea. This dataset, designed to comply with South Korea's Personal Information Protection Act (PIPA), avoids personally identifiable information while grounding AI agents in real demographics. It was created using NeMo Data Designer, an open-source system from NVIDIA that combines a Probabilistic Graphical Model with Gemma-4-31B for Korean-language narrative generation. The dataset is part of the broader Nemotron-Personas Collection, which includes similar resources for countries such as the USA, Japan, India, Singapore, Brazil, and France. NAVER Cloud contributed seed data and expertise during the design phase. A tutorial demonstrates how to deploy a Korean agent using this dataset in about 20 minutes via hosted APIs, with applications in domains like public health, finance, and education. The agent can be deployed using NVIDIA NemoClaw, NVIDIA NIM, or the NVIDIA API catalog. NVIDIA Nemotron Developer Days will be held in Seoul on April 21–22, 2026, featuring technical sessions and a hackathon focused on sovereign AI and open models.