Kumo develops foundation models specifically engineered for relational enterprise data housed within data warehouses. These models enable users to generate instantaneous, zero-shot predictions directly from structured data without requiring complex machine learning pipelines. The platform supports fine-tuning for custom needs while offering explainable outputs and reverse ETL capabilities for triggering real-time actions.
Funding
$36.6M raised to dateRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
Founders
Product
Problem
Traditional machine learning workflows require extensive manual feature engineering and the creation of training sets, which can be time-consuming and costly. Existing solutions often fail to fully leverage the relational structure of data stored in data warehouses, limiting the accuracy and efficiency of predictive models.
Solution
Kumo.AI provides a platform that leverages graph transformer architecture and pre-trained large language models to automate the generation of predictive models directly from raw relational data. By representing data as a graph and incorporating knowledge from LLMs, Kumo.AI eliminates the need for manual feature engineering, training set generation, and complex ML pipelines. The platform enables users to quickly deploy accurate predictive analytics for various use cases, including customer retention, fraud detection, and personalized recommendations. Kumo.AI integrates with data warehouses such as Databricks and Snowflake, allowing it to operate directly on raw data tables while utilizing Unity Catalog for data governance and security. The platform's automated machine learning pipelines keep models updated with the latest data, ensuring predictions remain accurate over time.
Target Audience
Kumo.AI targets data scientists, machine learning engineers, and business analysts who need to quickly build and deploy accurate predictive models from relational data without extensive manual effort.
Features
- Automated feature engineering using graph transformer networks and pre-trained large language models
- Direct integration with data warehouses like Databricks and Snowflake via Unity Catalog
- Support for various data types, including structured and unstructured data
- Automated machine learning pipelines for continuous model training and updating
- Ability to generate batch predictions or embeddings for downstream tasks
- Explainable AI features that offer insights into how predictions are made
- REST APIs for easy integration with existing systems
- SOC 2 Type II and GDPR compliance, with options for both SaaS and Private Cloud operating models