Kumo, a leader in predictive AI, has unveiled KumoRFM, the world’s first foundation model built for relational data, enabling businesses to generate accurate predictions—such as item recommendations, customer churn detection, and fraud identification—directly from their enterprise data in seconds and without manual effort. KumoRFM delivers 20x faster time-to-value and 30–50% higher accuracy compared to traditional approaches.
While AI has revolutionized how organizations handle unstructured data like text, images, audio, and video, structured enterprise data—such as customer records, transactions, and product catalogs—has largely remained underutilized in modern AI advancements. KumoRFM addresses this gap, bringing the capabilities of Transformer architecture to structured data at a fraction of the cost.
“To make predictions and business decisions, even the most advanced companies still rely on decades-old machine learning techniques for enterprise data,” said Jure Leskovec, Co-Founder and Chief Scientist at Kumo. “We’re proud to bring to enterprise data what GPTs brought to text.”
Unlike traditional models that require separate training for each predictive task, KumoRFM is a zero-shot model capable of delivering instant, task-agnostic predictions. It uses its understanding of traits, behaviors, and relationships in structured data to forecast business outcomes—powering decisions from fraud detection to personalized marketing offers, product recommendations, and ad targeting, all with direct revenue impact.
Data science teams, developers, and engineers can connect KumoRFM to their data warehouse via an API, enabling immediate pattern detection and prediction across multiple use cases in real time. This capability accelerates application development and deployment without the need for task-specific model building or training. Trained solely on synthetic enterprise-like data, KumoRFM is compact, cost-effective, and inference-ready.
“AI tools like chatbots and content generators have shown the power of language models, but enterprise data has been the missing piece—and KumoRFM fills that gap,” said Vanja Josifovski, Co-Founder and CEO at Kumo. “When AI connects with business data, that’s when real ROI and business impact happen.”
KumoRFM is a pre-trained Relational Graph Transformer model, capable of learning from multiple tables of structured enterprise data out of the box. Fine-tuning for specific tasks boosts its accuracy by 30–50% over traditional methods. Built on years of research in Graph Neural Networks (GNNs) and Graph Transformers—a field pioneered by Jure Leskovec and the Kumo team—this technology already powers Kumo’s platform, used by enterprises such as DoorDash, Databricks, Snowflake, and Reddit.
To learn more or get started, visit Kumo.AI. For more such updates, follow us on TechMediaGlobal.