Rasgo's feature store, powered by Snowflake, orchestrates feature engineering transformations programmatically for users directly within Snowflake’s data platform.
With the new functionality released by Snowflake to create Scala-native Snow SQL transformations and deploy Java UDFs, Rasgo's feature store can extend support for feature transformations from pre-existing complex libraries and open source standards. One such example is word stemming and lemmatization, which requires low-level string transformation to perform NLP on unstructured text data. These transformations take word phrases and distill them down to a standardized, grammar-less form that makes it easier for a model to recognize commonalities and patterns in the data. For example, word stemming of any entry in this list would result in the final form 'like': "likes","liked","likely","liking".
Rasgo orchestrates the creation and execution of the UDF to perform this and other NLP transformations, and facilitate training and inference data for NLP models directly from your Snowflake deployment.
