Kumu

Kumu is an innovative platform designed to make graph neural networks accessible to enterprises. It integrates seamlessly with existing data infrastructure, such as Snowflake and Databricks, allowing users to leverage their relational data without needing extensive graph learning expertise. The platform offers a user-friendly interface, including a SQL-like predictive query language, which simplifies the process of building and deploying advanced AI models.

Kumu's key features include automated machine learning (AutoML) capabilities, which enable users to predict future data points and automate time-consuming feature engineering tasks. By utilizing graph neural networks (GNNs), Kumu can learn complex features directly from data, eliminating the need for manual feature engineering. This approach supports a wide range of applications, from consumer behavior analysis to healthcare demand forecasting and fintech fraud detection.

The platform is designed to optimize infrastructure for real-world economics, combining CPU and GPU computing to achieve cost efficiency without sacrificing performance. Kumu also emphasizes trust and transparency, incorporating model explainability and bias detection into its core architecture. By integrating with established data warehouses, Kumu allows businesses to quickly derive value from their data without requiring significant changes to their existing systems.