Sayed Rafay Bin Shah

Researched and developed gradient monitoring algorithms to improve neural network performance in industrial ML applications. Designed transfer learning methods for handling challenges involving anomaly detection, predictive maintenance, and health forecasting. Worked collaboratively with industrial partners in VDI/VDE-funded ZIM projects to build and deploy data preprocessing pipelines and robust neural network models. Supervised Master theses on applied machine learning topics and achieved top results.

Conducted data analysis and feature engineering, and created machine learning models for fault detection in a ZIM project in collaboration with Bultmann GmbH.

Conducted quality assessments (stress, electrical, optical) on high-power electrical modules following automotive standards, along with anomaly detection via data analysis.

Researched gradient monitoring methods to reduce data drift and improve knowledge transfer in industrial ML applications by utilizing network parameter–gradient relationships. Developed unsupervised domain adaptation and continual learning frameworks for memory and computational-constrained environments and real-world use. Stabilized discriminator training in adversarial adaptation through gradient monitoring.