This Project explores the implementation of a computational model based on machine learning for the generation of synthetic melodies and harmonies, that might work as a base for the composition of musical pieces in rhythms that belong to the tradition and culture of the Colombian Caribbean Region. For the selection of the model a state-of-the-art analysis was performed. The model implemented corresponds to a stack of RNN layers, working upon the raw audio signal. This Project will work as a first attempt to the treatment of music from the Colombian Caribbean Region through machine learning and AI, its results show that it is possible to synthetize music from this region through machine learning modeling. Future work is proposed, which comprises the compilation of a bigger dataset and exploring the interaction between musical melodic instruments and percussion ones through the model.