Recommender systems aim at suggesting one or more next items to users, given a specific history of the user and/or relevant context. However, when we need to recommend an ordered set of items, the problem becomes increasingly complex. We call these ordered sets of items, bundles. Such bundles are, for example, a diet plan or a workout routine for a given time interval. Instead of one next meal or one next workout, we need to optimize for a bundle of items that the user will like over a given time interval (e.g., one month). Considering all possible combinations of items in such cases is prohibitively expensive due to the vast number of such combinations. In this paper, we build a recommender system that suggests bundles to users that cover their needs and preferences. While the state-of-the-art typically uses collaborative filtering (e.g., matrix factorization) or/and content-based filtering methods, we use Genetic Algorithms to efficiently explore the large space of all possible recommendations. We propose a fitness function that measures the suitability of each bundle and we perform an experimental evaluation on the recommendation of the diet plan using a food data set and an exercise data set.