CAPABILITIES
Unprecedented
time to value
Easy data preparation
Built-in connectors
One-click integrations to all commonly used enterprise data sources
Schema registration
No feature engineering pipelines or feature store. Just register your schema by linking tables together based on primary and foreign key relationships, and Kumo's Neural Networks take care of using the right features at the right time on your behalf
Data Quality
At every step of the data preparation process, enjoy comprehensive out of the box validations to check for data quality at the graph, table, and individual column level
Best in class model-building
Predictive Querying
In just a single sitting, use our purely declarative language to flexibly query the future of your enterprise in dozens of impactful ways in parallel, as easily as you use SQL to query the past today.
Each predictive query defines an end to end best practice GNN-based ML pipeline in just 2-3 lines
Customizable model planner
Either use our out of the box AutoML, or customize every aspect of the modeling process through Kumo’s in-depth advanced options.
Iterate and improve over time on top of Kumo’s out of the box baseline using your domain expertise.
Trust and Explainability
Model evaluation
Support for all industry standard charts and metrics for evaluating how well your model generalizes to future/unseen data
Explaining model behavior
Get detailed explanations of how your model behaves both at a population and an individual prediction level, enabling you to:
- Understand the importance of every column in your dataset to your end predictions
- Discover business-relevant trends and insights buried in your raw input data
- Understand exactly how different input values contributed to a specific prediction
- Build confidence that your model is doing the right thing!
Enterprise-readiness
Resilient production deployment
REST APIs backed by high availability SLAs covering all major features, enabling you to orchestrate programmatic workflows using Kumo.
Out of the box prediction and retraining monitoring/alerting functionality for detecting target drift, input data drift, model degradation, and other common production deployment issues
Security & Compliance
Available as single tenant SaaS where each customer has dedicated deployment of Kumo stack, preventing any data/model leakage between customers. No customer data persisted on Kumo servers after jobs completed.
Secure private deployments also available in your Databricks and Snowflake environments Full SOC 2 Type 2 compliance. Iterate and improve over time on top of Kumo’s out of the box baseline using your domain expertise.
Technology - Graph Neural Networks
(GNNs) out of the box’
Deep learning led to AI breakthroughs in image and language processing -- but, to date, deep learning methods have been largely limited in applicability to images (matrices) and text (sequences).
Your enterprise data naturally forms a graph, representing a rich, interconnected web of business relationships, interactions, customers, transactions, and more.
GNNs generalize deep learning to graph data. GNNs are the only deep learning architecture that is not fixed but adapts to the shape and structure of the underlying data.
Kumo brings GNNs to enterprise data in a way that is both scalable and easy to use. We represent it automatically and optimally in graph form, leveraging its inherent connectedness to deliver better accuracy even with less training data.
Single News: no article selected
Query the future
Unleash the predictive power of your enterprise data
Get Free Trial