TransmogrifAI
An AutoML library in Scala that runs on Apache Spark, automating feature engineering and model selection to build strongly typed, production machine learning pipelines with close to hand-tuned accuracy.
TransmogrifAI is an automated machine learning library written in Scala that runs on top of Apache Spark. It focuses on developer productivity, offering an API with compile-time type safety, modularity and reuse, so a machine learning pipeline can be assembled from typed, reusable components rather than hand-wired feature engineering code.
Through automation of algorithm and hyperparameter selection, it aims to reach accuracy close to a hand-tuned model with much less development time, letting a team build a production machine learning application without deep ML expertise. A worked example trains a model to predict Titanic survivors from the passenger manifest, and the library also supports fully specifying every feature and algorithm manually for users who want more control.
It is added to a project as a regular Spark dependency through Gradle or Maven, with released versions matched to specific Spark, Scala and Java version combinations. It is licensed under BSD-3-Clause and was developed at Salesforce, where it powers Einstein predictions.