Aerosolve by Airbnb

Aerosolve is a machine learning library that is focused on being human-friendly. It's meant to be used with sparse features such as keywords, filters, search terms, pricing, etc. It differs from other popular ML libraries because of its thrift-based feature representation: All features are clustered in logical groups (families of features). This enables expressing transformations on an entire feature family or different families of features. By transforming these, a new feature family can be created.

Other features are:

  • An intuitive feature transform language
  • Debuggable models
  • Scala for training
  • A distinct separate lightweight Java inference code.



Company Airbnb
Category Data Science
Subcategory ML Library
Availability Apache License 2.0
Earning Model Library
Ownership None
Stock None


Java ML Library model development Scala
Linkedin Employees
Linkedin Followers
Facebook Followers
Facebook Page Likers
Twitter Followers
Made with 🔨 by Roel Peters.