Kumo.AI, a leader in predictive AI, launched KumoRFM on May 20, 2025, introducing the world’s first Relational Foundation Model (RFM) designed to deliver instant, accurate predictions on structured enterprise data. This breakthrough eliminates the need for task-specific model training, offering 20x faster time-to-value and 30-50% higher accuracy compared to traditional methods.
KumoRFM delivers instant predictions on relational data with zero-shot capability.
Achieves 20x faster results and 30-50% higher accuracy than traditional models.
Connects via API to data warehouses for real-time business predictions.
Trained on synthetic data, ensuring cost-effective and compact inference.
Supports use cases like churn prediction, fraud detection, and recommendations.
Backed by $37M in funding from Sequoia Capital and enterprise clients like DoorDash.
KumoRFM addresses a critical gap in AI innovation by enabling instant predictions on structured enterprise data, such as customer records, transaction histories, and product catalogs. Unlike language models like ChatGPT, which focus on text, KumoRFM leverages its understanding of relationships within data warehouses to predict business outcomes like customer churn, fraud, or personalized recommendations. “AI tools like chatbots have shown what’s possible with language, but there’s a missing piece when it comes to enterprise data, and KumoRFM fills that gap,” said Vanja Josifovski, Co-Founder and CEO at Kumo.
Unlike traditional machine learning, which requires months to build task-specific models, KumoRFM is a pre-trained Relational Graph Transformer that delivers zero-shot predictions. Data teams can connect it to their data warehouse via an API, enabling immediate pattern recognition and predictions without manual training. “To make predictions, even cutting-edge companies are using 20-year-old techniques on enterprise data,” said Jure Leskovec, Co-Founder and Chief Scientist. KumoRFM’s zero-shot approach results in 30-50% higher accuracy and up to 10% additional improvement when fine-tuned.
Trained on synthetic enterprise-like data, KumoRFM is compact and cost-efficient, making it accessible to organizations of all sizes. Its Relational Graph Transformer architecture, built on years of Graph Neural Network research by Leskovec, processes complex multi-table data without feature engineering. The model supports diverse applications, including fraud detection, recommendation systems, and demand forecasting, and is used by clients like DoorDash, Databricks, and Reddit.
KumoRFM’s launch has garnered attention, with posts on X highlighting its ability to outperform traditional LLMs and specialized models in predictive accuracy. The model’s integration with platforms like Snowflake and Databricks enhances its scalability, operating on terabyte-sized datasets without data sampling. Kumo, founded by former Pinterest, Airbnb, and LinkedIn leaders, has raised $37M from Sequoia Capital and was named a 2024 A.I. Awards winner for Most Innovative AI Technology.
KumoRFM redefines predictive AI for enterprises by delivering instant, accurate insights from relational data. Its zero-shot capability, high accuracy, and seamless integration make it a game-changer for data-driven decision-making, driving real ROI across industries.
Kumo transforms how businesses create predictions from their data warehouse, using AI to quickly build high-performing machine learning models that help data scientists better predict user and customer behaviors with best-in-class accuracy. The company was founded by three PhDs who have held executive leadership and academic positions at Pinterest, Airbnb, LinkedIn, and Stanford. Kumo has raised $37 million in funding and is backed by Sequoia Capital. Kumo is reshaping the future of applications, making predictive AI accessible and practical for companies of all sizes.
TechIntelPro empowers B2B technology decision-makers with the latest industry trends, deep insights, leadership perspectives and actionable intelligence to drive both organizational and professional milestones—and thrive on the tech pulse.