Ciwan Ceylan

I am an industrial PhD student at the KTH Royal Institute of Technology and a member of the Wallenberg AI, Autonomous Systems and Software Program (WASP), supported by SEB, where I also work as a data scientist. My PhD is supervised by Prof. Danica Kragic, and co-supervised by Dr. Kambiz Ghoorchian at SEB, and Prof. Aristides Gionis at KTH.

My research focuses on unsupervised node embedding methods that are both mathematically analysable and computationally scalable, with applications to large-scale financial transaction networks. In particular, I develop models that capture complex graph properties such as multi-scale structures, edge directionality, and edge flow. These models enable the effective use of machine learning and data mining techniques on transaction networks and other large, structured datasets, without relying on costly labels, and while preserving the interpretability required for high-stakes, complex data analysis.

In the financial context, such models play an important role in anti-money laundering (AML) and counter-terrorism financing, where graph-informed machine learning systems can help identify suspicious behavior and support early detection of illicit activity. Long term, I aim to integrate node embeddings with techniques such as anomaly detection to build robust, explainable tools for combating financial crime.

Beyond transaction graphs, I explore the broader applicability of unsupervised methods to problems like node classification, graph alignment, link prediction, clustering, and network flow estimation. I’m also passionate about unsupervised learning beyond graphs, and my earlier work includes a paper on contrastive learning for image data.