My research focuses on the theoretical and practical exploration of Implicit Neural Networks and Randomly Wired Neural Networks. Unlike conventional layered architectures, these networks feature computational graphs that may contain directed or undirected cycles, offering new possibilities for modeling learning stability, memory, and self-organizing behavior. The aim is to develop cost-efficient yet high-performance models for time-series analysis and classification, combining the advantages of implicit architectures  and biologically inspired graph structures. Such networks exhibit small-world, modular, and hierarchical organization, aligning more closely with brain-like topologies.

Our research focuses on the development of intelligent, data-driven manufacturing systems, with special emphasis on the digitalization and automation of welding processes. We work on creating cyber-physical systems that enhance manufacturing efficiency and reliability through real-time sensor data and machine learning algorithms. Our main research areas include image-based welding defect detection, sensor-based measurement system development, and the combined use of finite element simulation and machine learning for structural optimization. In addition, we explore Fourier Neural Operators for time-series and acoustic signal analysis. The ultimate goal of our work is to develop digital and self-optimizing manufacturing environments that integrate artificial intelligence with real-time data processing.