Jorge Poco Research Statement

An important goal of my research is to design novel techniques and systems to support data exploration. Since real-world data is often complex, uncertain, and heterogeneous, to analyze and derive insights from these data requires extensive expertise in many disciplines. Thus, an important theme in my work has been on the integration of visualization-based approaches with techniques and algorithms from related areas such as machine learning, topology, and text processing, to improve and guide the data analysis process. The ever-increasing scale and complexity of data creates additional challenges. Data sets can not only be large, but analyses often require data from different sources and in different formats. To overcome these challenges, I have (i) devised novel visual encodings and visual analytics approaches to help domain experts generate insights out of complex data, and (ii) designed efficient algorithms that cut across multiple areas, such as, machine learning, topology and optimization, and aim to attain balance between the available resources and the ability to interactive analyze the data. During my doctoral studies, I was involved in two main projects: Multifaceted Climate Data Visualization, where we collaborated with climate scientists to develop data analytics techniques and visual representation to describe climate models; and Spatio-Temporal Urban Data Visualization, where we (1) proposed a new visual model to query origin-destination data and an efficient index structure to evaluate these queries at interactive rates, and (2) used techniques from vector field visualization to identify mobility patterns using New York City taxi data. During my masters, I worked on problems where I used multidimensional projection to explore High-Dimensional Data and proposed novel algorithms to accelerate the projections and allow users’ feedback to re-project the data interactively. These projects led not only to 15 publications but also to real systems that are currently being used by domain experts.