Context: Given the recent stunning advances of applying neural network (NN) models in the context of science, it is natural for PhD students to show an interest in using these models in their own research. Problem: However, in order for NN to be useful in research, one has to address the gap between the education about NN and the knowledge of how they can be applied. Objective: The goal of this thesis is to elucidate what PhD students need to learn in order to be able to use NN successfully and to develop a concept for a game-based learning framework that achieves this goal. Method: Through the use of a survey of Master and PhD students, we investigate what NN related competencies PhD students are interested in, what they need to learn and how they prefer to learn them. By then applying the resulting design principles and competencies to a concept for a learning game, we provide a prototype of this might look like. Finally, we provide an evaluation of this concept by interviewing representatives of the target group-PhD students. Result: We show that the concept that we developed was judged positively by the interviewees. The design principles and competencies that we extracted from the survey seem to be to the satisfaction of students. However, it remains unclear if a game can deliver the high information density that PhD students expect while learning, given that they are subject to strict time constraints. Conclusion: It appears as though the idea of a learning game that teaches NN holds merit and definitely can serve to increase the motivation of students that engage with this topic. Future work is needed in order to see if such a game, if fully implemented, truly can square with PhD students’ expectations or if it is better suited for undergraduate level-teaching.