Nowadays, scientific articles are mostly published as PDF files containing unstructured and semi-structured text. This way of scholarly communication severely limits the possibilities to automatically process and reuse scholarly knowledge. As a consequence, applying data analysis methods to scientific literature is a non-trivial process. FAIR scholarly knowledge graphs (SKGs) is one approach to represent scholarly knowledge in a structured, machineactionable, and semantic manner. In this thesis, we exploit SKGs, specifically the Open Research Knowledge Graph (ORKG), in data science. We introduce a generic architecture for applying data science to scholarly data. We then implement the architecture using the ORKG as the main data source and test it in two use cases in different domains. We demonstrate approaches that reuse the ORKG content to draw new insights and build applications and visualizations on top of SKGs data.