Machine learning researcher at Data Science Chair for Real-Time Decision Making.
Modern consumer products have become so complex that often, no single person has the knowledge necessary to troubleshoot them when they fail. Instead, customer service requires the coordination of several experts and has become increasingly costly, and nowadays cheaper products are often replaced rather than repaired. We believe there is a better, less wasteful solution: carefully recorded troubleshooting generates rich data that can be leveraged to automatize the detective work. We aim to leverage history of prior successful troubleshooting to assist representatives in solving those puzzles using machine learning and discrete optimization.