What Kumo does
Kumo is a predictive AI platform for enterprise relational data. Most generative AI today targets unstructured text and pixels, but the data that actually runs companies — customers, transactions, products, events — lives in interconnected SQL tables and warehouses. Kumo trains graph neural networks directly on top of that relational data, letting customers ask predictive questions in plain English ('which users will churn in the next 30 days?', 'which prospects will buy this product?') and get accurate answers without hand-built feature engineering or task-specific models.
In May 2025 Kumo released KumoRFM, billed as the world's first Relational Foundation Model — a foundation model for structured enterprise data that delivers instant predictions without task-specific training, claiming 20x faster time-to-value and 30-50% higher accuracy than traditional pipelines on benchmark predictive tasks. In 2025 Kumo also added MCP (Model Context Protocol) support so KumoRFM can be invoked directly from agentic systems.
Who it's for
Kumo targets data and ML teams at enterprises with rich relational data — fintechs, e-commerce, marketplaces, telcos, and SaaS — who want predictive AI without the months-long model lifecycle of bespoke ML.
Pricing
Kumo sells enterprise contracts based on data volume, prediction throughput, and deployment surface (Snowflake, Databricks, AWS). KumoRFM has a developer tier for evaluation but no public self-serve price card.
Team & funding
Kumo was founded in 2021 by Vanja Josifovski (CEO, former CTO of Airbnb Homes and Pinterest), Jure Leskovec (Chief Scientist, Stanford CS professor and one of the world's top graph-learning researchers), and Hema Raghavan (Head of Engineering, ex-LinkedIn). Kumo has raised approximately $37M total — a $19M Series A in 2022 and an $18M Series B in September 2022 led by Sequoia Capital, with angel backing from Frank Slootman, Sridhar Ramaswamy, Ben Silbermann, Matei Zaharia, Tristan Handy, Ron Conway, and others. The company is headquartered in Mountain View with engineering in Europe and roughly 50-70 employees.
Position vs competitors
Kumo competes with DataRobot, H2O.ai, Snowflake's Cortex, Databricks Mosaic, and bespoke in-house ML teams. Its KumoRFM and graph-native approach are unique versus the LLM-on-tables direction many incumbents are taking.