We present a collection recommender system that can automatically create andrecommend collections of items at a user level. Unlike regular recommendersystems, which output top-N relevant items, a collection recommender systemoutputs collections of items such that the items in the collections arerelevant to a user, and the items within a collection follow a specific theme.Our system builds on top of the user-item representations learnt by itemrecommender systems. We employ dimensionality reduction and clusteringtechniques along with intuitive heuristics to create collections with theirratings and titles. We test these ideas in a real-world setting of music recommendation, within apopular music streaming service. We find that there is a 2.3x increase inrecommendation-driven consumption when recommending collections over items.Further, it results in effective utilization of real estate and leads torecommending a more and diverse set of items. To our knowledge, these are firstof its kind experiments at such a large scale.