Model Storage allows you to store and manage trained AI/ML models directly in the CDP platform. By removing the need for external storage (e.g., Amazon S3), you can streamline AI deployment, improve performance, and simplify operations for real-time personalization.
Model Storage enables you to:
Store trained models in ONNX format and metadata directly in the platform.
Manage Models through the platform UI and APIs.
Retrieve models in real time for scoring, decisioning, and personalization.
Eliminate the need for external storage like AWS S3.
Note: This feature is available to users with the "Models" permission. To ensure access, please review your user role settings and update them if necessary.
Example use cases
Use Case | Description |
Engagement scoring | Score users in real time to predict churn, interest, or conversion likelihood. |
Look-alike modeling | Find users similar to high-value segments based on behavioral traits. |
Smart paywalling | Dynamically decide whether to show a paywall, discount, or offer. |
Discount optimization | Calculate ideal discount levels for maximizing revenue and conversions. |
Public API support
A public API is available to allow external systems to:
Upload models programmatically (e.g., from AWS, Azure, or local clusters).
Update models automatically when retraining occurs.
Integrate model workflows into DevOps pipelines.
This enables full model lifecycle management, from training to deployment, without manual steps.
Configure and manage models
Create a model
To start:
Select More > AI Workbench.
Click Models from the sub-menu.
Select Add model.
Enter a name.
Expand the metadata to add a description and label.
Select the upload icon to upload the ONNX file that contains your AI model.
Only ONNX file formats are accepted with a max size of 10MB.
Click Save.
Note: Versioning is not supported. Uploading a new model replaces the old one.
You can work with models using:
REST API: Create, update, retrieve, and delete models.
Python API: Optimized for use within AI Workbench notebooks.
Node.js API: For retrieving models in real-time use cases (read-only).
Model metadata
Each model includes key metadata for tracking and filtering:
Metadata Field | Example |
Model name |
|
Model type |
|
Features |
|
Parameters |
|