Kumo Unveils Relational Foundation Model for Instant Enterprise Predictions

Kumo has launched KumoRFM, the world's first Relational Foundation Model (RFM) designed to deliver instant predictions on enterprise data, announced in a press release. This innovative model allows businesses to generate accurate predictions, such as item recommendations and fraud detection, directly from their data without the need for manual model building and training.

KumoRFM is built on the Transformer architecture, extending its capabilities beyond natural language to structured enterprise data. The model provides a 20x faster time to value and achieves 30-50% higher accuracy compared to traditional machine learning techniques. It is designed to work out of the box, offering zero-shot predictions without task-specific training.

Data science teams can connect KumoRFM to their data warehouses via an API, enabling real-time predictions for various business-critical tasks. The model is trained on synthetic enterprise-like data, making it compact and cost-effective for inference. KumoRFM's core technology is based on years of research in Graph Neural Networks and Graph Transformers, pioneered by Kumo's team, and is already utilized by over 20 enterprise clients, including DoorDash and Snowflake.