The Abacus.AI platform allows you to process, join and transform raw tabular data present in various data sources (like AWS S3, Snowflake, Redshift, GCS, Salesforce, Marketo) into features for the purposes of building machine learning models and running predictions on them. Features are organized into Feature Groups and can be easily specified using ANSI-SQL or Python code. Specifying feature groups allows for a clear path to productionizing data pipelines that would further serve ML model training, batch (offline) predictions, and online serving (streaming) of predictions from a custom or Abacus.AI built model.