The Google framework called TensorFlow has been widely used for decision making in several areas, including Education. Predicting student risk and optimizing a student’s learning path are, for example, two traditional educational problems that have been explored for years but there are a myriad of different data mining approaches involved. This paper’s goal is to illustrate the results of a systematic mapping process conducted on educational data mining studies using the TensorFlow framework. Furthermore, this paper will assist in illustrating what kind of problems to focus on (which can paradoxically be seen as opportunities), identify, demonstrate, and catalogue all the academic studies that have discussed it, and the approaches leveraged (neural network, decision tree, natural language processing, and so on). The mapping process followed five phases with rigor, returning a set of 32 relevant papers in the study area with detailed information related to the research questions. The outcome of this systematic study will be of benefit to academic managers, researchers, and students who use this framework as support to solve educational problems. Keywords: systematic mapping; Educational problems; TensorFlow; data mining