The National Imagery Interpretability Rating Scale, or NIIRS, is an important measure ofinterpretation capability in aerial imagery. Many important ecological, aerospace, andsurveillance tasks depend on satellite data being interpretable to at least a certain NIIRS score,but it is rarely possible to acquire such high-resolution data experimentally. We hope toinvestigate the potential of Conditional Normalizing-Flows neural networks as a solution to thesatellite imagery-based super-resolution problem. This implementation requires developing adeterministic mapping between low resolution experimentally captured data and high-resolutiongenerated data automatically fit and labeled to satisfy NIIRS score requirements. Our model aimsto be capable of processing an image and its ground sampling distance to produce statisticallyinferenced predictions of the same scene at all other NIIRS scores.

Trained a scene-representation network called a Generative Query Network capable of producing new images of a certain scene given contextual information about the scene and a viewpoint from which to view the scene from. Performed analysis on the model to assess learning ability, product quality, and possible improvements.

Researched additive-manufacturing data management and analytics, including proposing a new standard file format for additive-manufacturing data based of the metadata centric HDF5 file format. This format was later linked to open-source pipelining tool DREAM.3D.Research was performed as a member of the Summer Undergraduate Research Fellowship.

Coach and private instructor for children aged three to fourteen. Taught basic water safety and comfort, beginner level technique for stroke, and advanced racing techniques for various competitive events. Assisted in meet management and facilitated events focused on ensuring the team environment was fun, spirited, and safe.Developed, introduced, and ran lifting and conditioning program for swimmers aged thirteen and up.