Automated Training and Inference Pipeline in Snowflake w/ Model Registry

Building an automated machine learning pipeline for streaming data can be a complex task, requiring data transformation, model training and hosting, and efficient prediction execution, often requiring services from multiple different providers.

Snowflake simplifies the process by offering features like Snowpark ML, streams, tasks, and the new Model Registry. I’d like to show you how easy it can be to construct an automated training and inference pipeline that can create predictions in parallel, entirely within the data cloud.

This pipeline provides the following benefits:

• Automated model retraining and deployment to Snowflake Model Registry
• Automated feature engineering and fast, parallelized predictions
• Familiarity and flexibility of python, using Snowpark
• Notification when training or predictions fail (Error notifications for tasks)
• Easy evaluation of model performance (Streamlit)

Click here to learn more!