Foundation models for structured and tabular data
Nums AI is a Seoul-based startup building foundation models for structured data. Its Causilo tabular foundation model predicts labels and values from examples in a table without task-specific retraining, and its Ephris graph foundation model performs in-context node classification. Models are available via API, cloud deployment, on-prem containers and open weights on Hugging Face with inference code on GitHub. In November 2026 Korean media reported a 2 billion won second pre-seed close, bringing its early funding to 6 billion won.