A scalable model for online contextual music recommendations

Engaging personalized recommendations are critical to the success of music streaming services. In this paper we experiment with a scalable design for building online personalized contextual recommenders across different music styles. We break down the architecture prominently into an online contextual recommender selection step and an online contextual content selection and ranking step. We discuss the value of this architecture by presenting experimental results on a genre-based personalized recommender on the home page of a global music streaming service.