In today's world software development is a competitive field. Being an expert gives software engineers opportunities to find better, higher-paying jobs. Recruiters are always searching for the right talent, but it is difficult to determine the expertise of a developer only from reviewing their resume. To solve this problem expertise detection algorithms are needed. A few problems arise when expertise is put into application: how can developer expertise be defined, measured, extracted or even learnt? Our work is attempting to provide recruiters a data-driven alternative to reading the candidate's CV or resume. In this thesis, we propose three novel topic modeling based, robust, data-driven techniques for expertise learning. Our extensive analysis of cross-platform developer expertise suggests that using multiple collaborative platforms is the optimal path towards gaining more knowledge and becoming an expert, as cross-platform expertise tends to be more diverse, thus creating opportunities for more effective learning by collaboration.