People love listening to music and so do we. Today, music streams are readily available through services such as YouTube, Spotify and Apple Music. The digitization of music has eased the way for users to gain access to large collections of music through the Internet. With all this music around us, it would be expected that music discovery is a settled field of research when in fact it is quite the contrary. Novel music discovery remains a challenging and a tedious task for users. Existing techniques such as content-based and collaborative filtering approaches recommend new music based on the past listening history of users. While there are obvious merits to such approaches, they also have some serious limitations with regards to music discovery. By focusing on the past, it becomes increasingly difficult for users to discover music that is different from what they have listened to in the past. Using accuracy as an evaluation metric, as the system become more `accurate' at giving you the same kind of music recommendations, serendipity and novelty often take a backseat. Cold-start is a well-known challenge that affects content-based and collaborative-filtering based music recommender systems. When a new user joins the system, because they have not yet rated or listened to anything yet, there is very little available information based on which any credible recommendations can be made. Lastly, traditional approaches to music recommender systems do not take into account the user's surroundings for music recommendations---thereby missing out on an opportunity to overcome the challenges put forth above. Our goal for this thesis is to design, implement and evaluate a learning-free novel music recommender system that allows users to easily discover new music based on their current surroundings.