Nums AI, a Seoul-based artificial intelligence startup developing foundation models for structured data, has secured ₩2 billion (approximately $1.5 million) in a second pre-seed funding close, bringing its total pre-seed financing to ₩6 billion (approximately $4.48 million).
The latest investment came from Mirae Asset Venture Investment and Korea Investment Partners, with each firm contributing ₩1 billion. The financing follows an initial ₩4 billion pre-seed investment completed in July 2026, backed by Stonebridge Ventures, SBVA, KT Investment and Base Ventures.
Founded in February 2026 by a research team led by Jaemin Yoo, a computer science professor at Seoul National University, Nums AI develops foundation models designed to process structured datasets and generate predictions without requiring users to train separate models for individual tasks.
Its primary model, Causilo, works with tabular data, including transaction records, customer information and operational datasets. It uses previously labelled examples to perform classification and regression tasks, supporting applications such as customer churn prediction, sales forecasting and credit-risk analysis.
The company has also developed Ephris, a graph foundation model that uses relationships between connected data points to predict unknown labels. Unlike Causilo’s focus on tabular datasets, Ephris is designed for graph-structured information.
Nums AI intends to use the additional capital to accelerate model development and recruit technical talent.
The startup was also selected for South Korea’s Technology Incubator Program for Startups (TIPS) R&D programme, securing ₩800 million in research funding separate from its venture financing.

