"Recommender Systems have become essential personalized navigational tools for usersto wade through the plethora of online content as they allow users todiscover relevant information that they would have never known itexisted. In recent years, the importance of this information discoveryprocess as opposed to explicit (keyword-based) search has been emphasized.Current research in Recommender Systems, while taking into account therelation between user and item, often ignores the ``context'' of therecommendation. We define as ``context'' any environmental, temporalor otherwise variable that influences a decision a user might make.Early work on Context-Aware Recommender Systems (CARS) has found thatcontextual factors do influence the recommendation needs of users.However, the role that each of the contextual variables (e.g. time,location, activity, emotional state, social network, etc.) plays onthe user's needs is still not clearly defined.The main aim of this proposal is to build a compact Context-AwareRecommender System (CARS) for mobile and desktop computing devices.The research methodology of this proposal is structured in 3 researchobjectives:1) Understanding contextual information in Recommender SystemsWhere data will be mined in order to uncover underlyingpatterns in the influence of context on users' preferences.2) Building Context-aware Recommendation modelsWhich involves using state of the art Machine Learning to buildmodels and algorithms for CARS3) Building a prototype and deploymentWhich involves building and deploying a prototype based on thedeveloped algorithms and conducting a user studyModern Machine Learning algorithms have been shown to perform well inRecommendation Tasks and this proposal has a strong algorithmic andmethods focus but also aims at knowledge discovery both through DataMining and Human Computer Interaction techniques."