PhD in Time Series Representation Learning for Industrial Process Characterization

This PhD position at KU Leuven’s Bruges Campus focuses on advancing time series representation learning for complex industrial processes. The interdisciplinary M-Group, combining expertise in computer science, electrical engineering, and mechanical engineering, seeks a candidate to design scalable, explainable algorithms for mining multivariate time series data. The project aims to create holistic process models using motifs—representative subsequences—that balance computational efficiency with interpretability. The successful applicant will collaborate with industrial partners, validate methods on real-world datasets, and contribute to both academic publications and industry applications. The position offers a fully funded 4-year scholarship within a collaborative, internationally oriented research environment.

• Design and implement novel algorithms for mining motifs in multivariate time series data to create explainable representations of industrial processes.
• Develop scalable methods that balance computational efficiency with representativeness of process dynamics.
• Incorporate weak supervision techniques to enable domain experts to guide algorithmic outputs without exhaustive labeling.
• Validate proposed methods on real-world industrial datasets provided by project partners, focusing on tasks like anomaly detection and root cause analysis.
• Collaborate with interdisciplinary teams to integrate findings into broader industrial process optimization frameworks.
• Present research results at international conferences and publish high-quality papers in top-tier venues.
• Engage with industrial stakeholders to translate academic research into actionable insights for production improvements.