Skip to navigation
About Kumo Predict

Kumo Predict Overview

View as Markdown
Coming soon

Kumo containers will be publicly available in the NVIDIA NGC Registry soon. Stay tuned.

The Kumo Predict Platform, also known as Kumo Studio, is Kumo’s managed web interface for training, evaluating, and deploying graph neural network models on relational data - no code required. Connect your data warehouse, define predictive queries, train models, and generate batch predictions, all from Kumo Studio.

For SDK-based workflows, see the Kumo Predict SDK instead.

Connect Data

Connect to S3, GCS, ADLS, BigQuery, Snowflake, or Databricks.

Train Model

Write predictive queries, configure model settings, and train.

Run Models

Generate batch predictions and export results.

References

Manage quotas, security, deployment options, and SSO.

Get Started

Review Requirements

Confirm platform requirements and choose the right GPU resources for your workload.

Make Your First Prediction

Connect data, create a graph, define a predictive query, and train a model in the platform.

Explore Examples and Solutions

Explore predictive-query examples and solution workflows for common business problems.

Reference

Look up the Kumo Predict SDK and its API reference.