Kumo.AI
Kumo.AI builds a relational foundation model (KumoRFM) that delivers zero-shot and fine-tuned predictive AI directly on enterprise data warehouses. Its platform serves data scientists, ML engineers, and CXOs at large enterprises including Walmart, SAP, DoorDash, Reddit, and Snowflake.
- Company typePrivate
- Founded2022
- HeadquartersMountain View, United States
- Headcount51–100
- GTM typeB2B
- OfferingSoftware
What Kumo.AI does
Kumo.AI was a Mountain View, California-based enterprise SaaS company founded in 2022 by Stanford researchers Vanja Josifovski (ex-CTO Airbnb and Pinterest), Jure Leskovec (Stanford professor and co-creator of Relational Deep Learning and Graph Neural Networks), and Hema Raghavan (ex-Sr. Director of Engineering, LinkedIn). The company built a predictive AI platform whose centerpiece was KumoRFM, a relational foundation model — built on a proprietary Relational Graph Transformer (RelGT) architecture — that performs in-context predictions directly on multi-table enterprise data warehouses without task-specific training or feature engineering. Users express predictions through a SQL-like declarative interface called Predictive Query Language (PQL) or a Python SDK (kumoai), with deployment modes spanning SaaS, customer VPC, Private Link, Databricks Native Application, and Snowflake Native App.
The product surface extended to KumoRFM-2 (the second-generation foundation model with hierarchical attention, scaling to 500B+ rows and outperforming supervised ML across 41 enterprise benchmarks), KumoRFM Fine-Tuning, Kumo Online Serving (a two-stage train-then-distill architecture delivering sub-100ms p99 latency at thousands of QPS), and the open-source Kumo Coding Agent Skills library that integrates with Claude Code and OpenAI Codex. Go-to-market combined an enterprise field-sales motion ("Book a Demo," forward-deployed engineer support, research agent, fine-tune) with a product-led free-tier funnel via kumorfm.ai, marketplace listings on Databricks and Snowflake, and a community-led developer motion (Discord, GitHub, arXiv/ICLR publications). Named customers included Walmart, SAP, DoorDash, Reddit, Snowflake, Databricks, Expedia, iFood, and Sainsbury's across 15 industry verticals. Total venture capital raised was approximately $37 million, led by Sequoia Capital's $18.5M Series A in April 2022. In June 2026, NVIDIA Corporation acquired Kumo.AI in a deal reportedly valued at over $400 million, with the three co-founders joining NVIDIA.
Kumo.AI firmographics
Firmographics- Name
- Kumo.AI
- Legal name
- Kumo AI, Inc.
- Website
- https://kumo.ai
- Company type
- Private
- Founded year
- 2022
- Operating status
- Acquired
- Headcount range
- 51–100 employees
- Short description
- Kumo.AI builds a relational foundation model (KumoRFM) that delivers zero-shot and fine-tuned predictive AI directly on enterprise data warehouses. Its platform serves data scientists, ML engineers, and CXOs at large enterprises including Walmart, SAP, DoorDash, Reddit, and Snowflake.
- Ownership category
- akta.pro rank
Kumo.AI industry classification
Industry- Product category
- Enterprise Predictive AI Software
- NAICS
- Software Publishers (5132)
- SIC
- Services-Prepackaged Software (7372)
- akta.pro primary industry
- AI Observability, Monitoring & Evaluation Platforms (Drift, Quality, Safety) (HDAEANAF)
- akta.pro secondary industry
- Audit, Explainability & Accountability Tooling (traceability, reporting) (HDAAAKAL)
Keywords
Where Kumo.AI is headquartered
LocationHeadquarters
- HQ city
- Mountain View
- HQ country
- United States
- HQ region
- North America
Markets served
Kumo.AI business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Technology or R&D, Personnel, Infrastructure, Marketing or Sales, Operations
Revenue model
- SaaS subscription: Primary revenue stream: managed SaaS platform with Apache Spark-based data platform, expanded choice of supported data warehouses (Amazon Redshift, AWS S3, Google Cloud BigQuery, Databricks, Snowflake), earlier feature access, quicker bug fixes, and enterprise support tier. Pricing not publicly disclosed; customers 'Book a Demo' to engage.
- Virtual Private Cloud / Private Link enterprise license: Enterprise-grade revenue from VPC and Private Link deployments where Kumo runs as a self-contained Kubernetes deployment inside the customer's VPC/VNet with customer-managed GPU and high-memory nodes, or as a Kumo-managed private environment accessible via Private Link, Private Service Connect, or IP allowlist.
- Data warehouse native app licensing: Revenue from native applications on Databricks (running on customer's Databricks Spark with Kumo-owned compute) and Snowflake Native App (running in Snowpark Container Services in customer's Snowflake account).
- Free tier / freemium API access: Free KumoRFM API key and self-serve onboarding (kumorfm.ai) intended as a top-of-funnel land motion for the SDK, with code-agent skills and free notebooks lowering the activation barrier before paid enterprise conversion.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Other | Annual | Enterprise SaaS (quote-based) |
| Freemium | Pay-as-you-go | Free tier / API key |
| Subscription | Multi-year contract | VPC / Private Link enterprise deployment |
Go-to-market motion5 records
Kumo.AI product offering
Product offeringCore offering
Kumo.AI provides a SaaS predictive AI platform built around KumoRFM, a relational foundation model that performs in-context predictions directly on enterprise relational data warehouses without task-specific training or feature engineering. Users express predictions in a few lines of Predictive Query Language (PQL), and the platform handles feature engineering, model optimization, and deployment across fraud, churn, demand forecasting, recommendation, and lead scoring use cases.
Product overview
Kumo.AI offers a unified predictive AI platform for relational business data, built around a flagship relational foundation model called KumoRFM (with KumoRFM-2 as its latest major version). The platform combines the foundation model with KumoRFM Fine-Tuning for task-specific optimization, Kumo Online Serving for sub-100ms real-time inference, the open-source Kumo Coding Agent Skills for natural-language-driven model building via Claude Code and OpenAI Codex, and a SQL-like Predictive Query Language (PQL) and Python SDK (kumoai) for declarative predictions. The core platform is delivered as a SaaS product, with additional deployment options including Virtual Private Cloud, Private Link/Private Service Connect, Snowflake Native App, and Databricks Native Application — letting enterprises run Kumo within their own cloud, VPC, or data warehouse boundary.
Differentiator
Problem solved
Functional benefit
Brands
- KumoRFM: Relational Foundation Model, the flagship AI platform and foundation model for predictive AI on enterprise relational data.
- KumoRFM-2
- Kumo Online Serving
- Kumo Coding Agent Skills
Products and services
- KumoRFM
- KumoRFM-2
- KumoRFM Fine-Tuning
- Kumo Online Serving
- Kumo Coding Agent Skills
- Predictive Query Language (PQL)
- Kumo Python SDK (kumoai)
- Kumo Platform (SaaS)
- Virtual Private Cloud (VPC) Deployment
- Kumo Private Link Deployment
- Snowflake Native App
- Databricks Native Application
Companies that use Kumo.AI
Customer profileIdeal customer profiles2 records
Kumo.AI technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
Integration14 records
AI capability13 records
Feature8 records
Kumo.AI partnerships and signals
Strategic signalRecent moves6 records
Expansion highlights6 records
Kumo.AI competitors and assessment
Company assessmentDirect peers
- DataRobot: Enterprise AI/ML platform offering automated predictive model building, deployment, and governance. Most direct competitor to Kumo's SaaS predictive-AI offering for Fortune 500 buyers, with overlapping use cases in churn, fraud, and demand forecasting.
- H2O.ai: Provider of H2O Driverless AI and the open-source H2O platform for automated machine learning on tabular and relational enterprise data. Competes head-to-head with KumoRFM's promise of high-accuracy predictive modeling without manual feature engineering.
- Palantir Foundry: Enterprise data and AI operations platform with deep ontology modeling and predictive analytics on relational enterprise data, deployed at large Fortune 500 customers. Competes for the same enterprise predictive-AI budget and the same buyer persona (CXO and data science leadership).
Broad incumbents
- Databricks Lakehouse AI: Databricks' native ML and AI capabilities (including MosaicML and Lakehouse AI) bundle predictive modeling, feature store, and governance into a single lakehouse platform. Both a partner (Databricks Native App) and the most capable direct competitor to Kumo on Databricks-resident data.
- Snowflake Cortex AI: Snowflake's built-in ML and LLM functions inside the data cloud. Kumo runs as a Snowflake Native App but Snowflake Cortex increasingly offers in-platform predictive and generative AI, making it a key distribution partner and a long-term competitive threat.
- Google Vertex AI: Google's end-to-end ML platform offering AutoML, Tabular Workflows, and foundation-model customization directly on BigQuery — the same data warehouse Kumo supports. Bundles predictive modeling, MLOps, and GenAI into a hyperscaler platform.
- AWS SageMaker: Amazon's flagship ML platform covering data prep, training, AutoML, and deployment natively on S3 and Redshift — the same primary data sources Kumo targets. Represents a default incumbent that enterprises often choose without evaluating specialized alternatives.
- Salesforce Einstein: Predictive AI layer inside the Salesforce CRM platform for lead scoring, churn, and forecasting on customer data. Overlaps with several of Kumo's flagship use cases and is sold through an enterprise field-sales motion to similar CXO buyers.
Emerging players
- Tecton: Enterprise feature platform that powers real-time and batch ML feature engineering on relational warehouse data. Adjacent rather than directly overlapping, but Tecton is frequently the underlying feature layer for the predictive models Kumo aims to replace with in-context learning.
- Weights & Biases: Developer-focused MLOps and experiment-tracking platform increasingly extending into model evaluation and AI observability. Comparable as an enterprise AI tooling brand targeting data scientists and ML engineers, and a relevant comparison point for Kumo's developer-led GTM motion.
Market position
Strengths5 records
Weaknesses4 records
Competitive moat5 records
Key risks1 record
Key highlights7 records
Customer concentration
Kumo.AI social profiles
Digital presenceKumo.AI compliance and trust
Trust signalCompliance7 records
Kumo.AI financial estimates
Financial estimateRevenue estimate
Valuation estimate
Kumo.AI leadership team
Management profileNumber of profiles
Profiles5 records
Kumo.AI subsidiaries and ownership
Company hierarchySubsidiaries1 record
Kumo.AI funding detail
Funding detailFunding overview
Funding rounds2 records
Investors4 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Kumo.AI M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about Kumo.AI
What does Kumo.AI do?
Kumo.AI provides a SaaS predictive AI platform built around KumoRFM, a relational foundation model that performs in-context predictions directly on enterprise relational data warehouses without task-specific training or feature engineering. Users express predictions in a few lines of Predictive Query Language (PQL), and the platform handles feature engineering, model optimization, and deployment across fraud, churn, demand forecasting, recommendation, and lead scoring use cases.
Is Kumo.AI a public or private company?
Kumo.AI is a private company. It is classified as corporate owned and is currently acquired.
When was Kumo.AI founded?
Kumo.AI was founded in 2022. It employs 51 to 100 people.
Where is Kumo.AI based?
Kumo.AI is headquartered in Mountain View, United States, in the North America region.
How does Kumo.AI make money?
Four revenue lines are on record. SaaS subscription is the primary driver. The others are virtual Private Cloud / Private Link enterprise license, data warehouse native app licensing and free tier / freemium API access.
Who are Kumo.AI's main competitors?
Direct peers on record are DataRobot, H2O.ai and Palantir Foundry. Broad incumbents are Databricks Lakehouse AI, Snowflake Cortex AI, Google Vertex AI, AWS SageMaker and Salesforce Einstein. Emerging players are Tecton and Weights & Biases.
Does Kumo.AI have an API?
Yes. Kumo offers a Python SDK (kumoai) and a REST API with a single authenticated endpoint for making predictions, training models, and managing PQL queries. SDK is installed via pip and authenticated via KUMO_API_KEY. The platform exposes its predictions and graph engine through authenticated APIs, with online serving delivered as a JSON request/response endpoint at sub-100ms latency. Also exposes an open-source coding agent skill set (kumo-coding-agent) that integrates with Claude Code and OpenAI Codex to enable natural-language-to-PQL workflows. Developer documentation is at kumo.ai/docs/rfm/overview.
What industry is Kumo.AI in?
Kumo.AI's product category is Enterprise Predictive AI Software. Its primary akta.pro industry code is HDAEANAF, AI Observability, Monitoring & Evaluation Platforms (Drift, Quality, Safety), with a secondary code of HDAAAKAL, Audit, Explainability & Accountability Tooling (traceability, reporting). Its NAICS code is 5132 and its SIC code is 7372.