Konstantin Schekotihin

Various AI applications in Failure Analysis (FA) have already shown that many routine tasks can be successfully automated. These tasks include identifying physical failures in images, labeling job reports, recommending analysis tasks, and retrieving important textual or visual information. However, most case studies tend to focus on a single application within traditional data science contexts, which typically involve the collection of a dataset, its labeling, model training, and deployment. This approach is effective as long as the number of deployed models remains small and can be managed by a limited group of FA engineers. In reality, an FA lab may deal with a large number of physical faults, often in the hundreds, making the conventional data science approach impractical. This talk will explore potential strategies for integrating AI into the workflows of an FA lab aimed at ensuring the stable and sustainable development and operation of AI components.

Konstantin Schekotihin is an associate professor at the University Klagenfurt, focusing on hybrid AI systems that integrate deep learning for computer vision and natural language processing with symbolic AI methods, like ontologies. His research has been applied in industrial projects involving failure analysis, cyber-physical systems, planning, and recommendations. He has authored over 90 papers in top AI conferences and journals, and has been recognized with six best paper awards. In the FA area, Konstantin publishes in prestigious venues like IEEE IPFA and ISTFA, covering AI approaches for automating failure analysis routines and AI infrastructures for FA labs.