Kumo has unveiled KumoRFM-2, a state-of-the-art relational foundation model that aims to change the way enterprises derive predictive insights from structured data. It is a big step forward from the traditional machine learning methods which involve the flattening of multi-table datasets and the loss of critical relational context. The newly developed model is based on a Relational Graph Transformer architecture which allows it to work directly on interconnected data tables and to keep track of relationships across datasets this is a major step in overcoming the limitation of enterprise AI systems that has been around for a long time. While older models and even current large language models convert relational data into single tables, KumoRFM-2 treats data as a graph, thus enabling it to provide more accurate and context-sensitive predictions on a large scale. The platform demonstrates exceptional performance capabilities, processing data at speeds of up to 5 GB per second with 20 million lookups per second, while scaling to datasets exceeding 500 billion rows, making it suitable for large, production-grade enterprise environments. “Enterprise data – customer records, transactions, product catalogs – holds enormous untapped revenue potential. Until now, using that data to generate business predictions required months of feature engineering and deep data science expertise, putting it out of reach for most teams,” said Dr. Vanja Josifovski, Co-Founder and CEO at Kumo. “KumoRFM-2 changes that: it’s the only model that actually understands the relationships across your tables instead of destroying them, it scales to hundreds of billions of rows, and it lets any team ask predictive questions in natural language. No feature engineering. No data science expertise required.”
Also Read: Vendavo Strengthens B2B Pricing with AI-Powered Assistants and ML-Driven Rule Generation
The model integrates directly with SQL databases and major cloud data platforms such as Snowflake, Databricks, and Apache Spark, pushing computation closer to the data layer for improved efficiency and real-time inference. Additionally, KumoRFM-2 introduces a hierarchical attention mechanism that enhances how it processes data across rows, columns, and relationships, improving prediction accuracy across diverse enterprise use cases. “For years, AI has been constrained by a fundamental limitation of not being able to reason over structured enterprise data. Database is not a document, it is a graph of relationships,” said Dr. Jure Leskovec, Co-Founder and Chief Scientist at Kumo. “KumoRFM-2 is the first model that sees the full graph. We developed Relational Graph Transformers, where the AI model can attend to any datapoint, preserving the complete structure of relational data at arbitrary scale.” Benchmark tests clearly illustrate that KumoRFM-2 method does better than top supervised machine learning models and tabular AI systems on various datasets and from different industries, at the same time, lowering the demand for large training data and still showing very good performance even when the data is noisy or to some extent incomplete. By doing away with complicated feature engineering and allowing predictive queries through natural language, KumoRFM-2 makes it possible for everyone to use advanced AI features. This is giving companies the power to speed up making decisions, find new value in their data, and launching predictive applications at a faster and more efficient rate than ever before.
