Kumo AI

Overview

Kumo AI is building a relational foundation model (KumoRFM) that operates directly on relational databases and graph-structured data. Rather than flattening tables into single dataframes, Kumo preserves the relational structure to learn richer representations across interconnected tables.

Key Product

KumoRFM is a relational foundation model that ingests multi-table relational databases and generates predictions across entities without manual feature engineering. KumoRFM uses graph neural network architectures to capture relationships between tables, targeting enterprise use cases like customer churn, fraud detection, and demand forecasting.

Team

Founded by leading researchers in graph ML and database systems, including faculty from Stanford. The team combines deep academic expertise with enterprise software experience.

Outlook

Kumo AI’s relational approach addresses a genuine gap: most tabular models assume single-table input, while enterprise data lives in multi-table schemas. Their $37M Series A and strong technical team position them well, though adoption depends on enterprises being willing to grant model access to their relational databases.