This PhD position at the Korteweg-de Vries Institute for Mathematics, University of Amsterdam, focuses on the mathematical foundations for explainable artificial intelligence (AI). The successful candidate will join the stochastics group and work on the NWO-funded project 'Mathematical Foundations for Explainable AI,' supervised by Dr. Tim van Erven. The project aims to develop new mathematical techniques and frameworks that make AI models more interpretable, generating understandable explanations for decisions made by machine learning systems. This research is motivated by the need for transparency in AI, such as understanding why a model predicts certain outcomes in healthcare or finance. The work will involve both theoretical development—proving mathematical guarantees for new explanation methods—and practical experimentation, including programming in Python to test these methods. The PhD student will initially select a pre-specified subproject to build expertise, then gradually shape their own research directions. Collaboration with other team members and international partners is expected. Responsibilities include conducting research, publishing in academic journals and conferences, participating in seminars, engaging in local and international collaborations, and contributing to teaching and supervision of undergraduate or Master students. The position offers a four-year contract (initially 18 months, extendable upon satisfactory evaluation), with a gross monthly salary ranging from €3,059 to €3,881, plus holiday and end-of-year allowances. Applicants must have a Master’s degree in mathematics or AI/computer science with a strong mathematical component, excellent programming skills, and strong communication abilities. The KdV Institute provides a vibrant research environment with active collaborations and participation in national research clusters. Applications are accepted via the Ellis PhD program until October 31, 2025. Interested candidates should apply through the Ellis PhD program application form and not via email or the FindAPhD page.