Graphs, a general type of data structures for capturing interconnected objects, are ubiquitous in a variety of disciplines and domains ranging from computational social science, recommender systems, medicine, bioinformatics to chemistry. Representative examples of real-world graphs include social networks, user-item networks, and protein-protein interaction networks. In this talk, I will introduce our work on learning effective representations of graphs such as learning low-dimensional node representations (the LINE paper, WWW’15), learning extremely low-dimensional node representations for graph visualization (the LargeVis paper, WWW’16), and our recent work on knowledge graph embedding (the RotatE paper, ICLR’19). Finally, I will introduce a new system GraphVite, which is specifically designed for learning representations for large-scale graphs ( WWW’19).