AI for real machines : bridging the research-industry gap

We interact with AI daily: writing emails, navigating traffic, or receiving personalized recommendations. AI is no longer just the future - it is the present. At the same time, industry is undergoing its own transformation, striving to embed AI into manufacturing, services, and business processes. But how far along are companies really? Have they even completed the previous industrial revolution? Bringing cutting-edge AI research into real-world industrial environments reveals a number of persistent gaps. In this talk, I will present examples of such gaps and share practical AI use cases from collaborative research projects with industry partners, aimed at anomaly detection, predictive maintenance, product quality monitoring, and automatic sensor calibration. I will highlight key challenges that are often encountered, such as limited and noisy data, scalability of deployment across diverse systems, high variability in machine behaviour and operating conditions, uncertainty quantification, and the need for transparent, interpretable AI models. I will also discuss how combining AI techniques with physics-based models and domain expertise enables practical, robust, and trustworthy solutions—demonstrating AI that truly works for real machines.