NEA is a tool for compressing the spatial footprint of an ensemble dataset. Ensemble datasets are formed from running a simulation multiple times, usually with a small change in initial parameters. This is useful for finding agreement and uncertainty. By taking advantage of the data similarity across runs and applying known dimensionality reduction algorithms, NEA can reduce ensembles by 7x or more. This work was published in EGPGV 2017's proceedings.
Intel's OSPRay and SWR renderers are designed for high performance rendering on multi- and manycore processors (including KNL). I am involved with integrating their use in the VisIt visualization application. Instructions on downloading a prebuilt binary or compiling the suite yourself are available at the link below. This work has been presented at Supercomputing 2015, 2016, and 2017.
PGen World is my final project for a computer graphics class in Fall 2015. It uses 2D Perlin noise to generate the surface of a planet with an atmosphere and two moons. The planet reflects the stars in its oceans. The moons orbit the planet and you can rotate the system with the mouse. Refresh to regenerate the textures.
TimeMiner is a final project from our Digital Archaeology course in Fall 2014. It was one of the first projects I worked on as a graduate student. By providing a "seed" article, we analyzed word usage to generate a group of articles by topic. Then we performed a temporal analysis of edits to the group to track when important events had occurred. We visualized the frequency of edits for the group as a streamgraph. Below are three groups that were analyzed.