Developing computer vision algorithms for autonomous robots in industrial automation

Researched and developed novel machine learning algorithms. Defined and led multiple machine learning and computer vision projects; served as machine learning lead on multi-disciplinary projects. Communicated technical concepts to stakeholders with varying degrees of technical expertise (researchers, managers, clients). Hired interns and supervised multiple successful intern projects resulting in publications and patents.Sample projects:* Deep synthesis of medical imagesLed computer vision research on a project studying the work practices of human radiologists. Used a deep network to synthesize artificial but realistic medical images with controlled parameters (background, number and type of lesions, etc.). This enabled studying human decisions in controlled conditions, greatly increasing the statistical power of experiments.* Anomaly and fraud detection in credit card transactionsDeveloped a novel probabilistic graphical model that accepts a stream of heterogeneous events and automatically separates them by source or purpose. The model was applied to a stream of credit card transactions for fraud detection, as well as for characterizing user behavior. The same model was subsequently also used to analyze a stream of user activity at a computer to detect insider threats. * Microchip transfer optimizationLed optimization work for the hardware project of transferring millions of microchips between multiple physical substrates. Gathered requirements and investigated hardware constraints. Proved the problem is NP-hard and developed heuristics that allow efficient optimization in typical cases. Improved algorithm runtime and efficiency to enable practical industrial application of the technology.* Information extraction from tables and receiptsLed development of novel computer vision algorithms for extracting information from document images. Several were transferred to a facility processing over 1 million health insurance claim forms per month.

Explored applications of nonparametric probabilistic graphical models in computer vision and recommendation systems. Published multiple papers, with two collecting over 100 citations each.