Onboarding (i.e., the process of incorporating new people) is relevant because it introduces employees to their role, the company’s culture, and what the company has to offer. Onboarding is then dependent on the company’s culture and practices. When it comes to software development, these practices include the methods, the tools or the developers’ organigram. Accordingly, there is not a one-size-fits-all onboarding, rather this procedure needs to be tuned for the practice at hand. This work tackles the specifics brought about by Software Product Line Engineering w.r.t. traditional software development, namely: larger code base, larger code variability, and larger and more heterogeneous teams. Specifically, this works advocates for feature-centric onboarding. Features (i.e., functional characteristics that are visible for a user) already play a key role throughout the SPL lifecycle. In this context, we advocate for defining the onboarding process as a journey where milestones are equated with features. Unfortunately, finding the most appropriate feature for a newcomer, if conducted manually by mentors, would be time-consuming, given the sheer number of features. To face this problem, we advocate for Recommender Systems based on the similarity between the feature’s codebase and the code previously explored by the newcomer. To this end, we resort to Topic Modeling, and specifically, Latent Dirichlet Allocation. We provide proof-of-concept through RecomMentor, a recommender system for pure-variants as the variability management system. RecomMentor is put to test against ranking metrics of the Information Retrieval literature. The first evaluation suggests that LDA could be an appropriate technique, paving the way towards using Recommender Systems in feature-based onboarding scenarios.