PhD Position in Graph Signal Processing, Machine Learning, and Geometric Deep Learning at Télécom Paris

A fully funded PhD position is available at Télécom Paris (Institut Polytechnique de Paris) within the IMT Futur, Ruptures & Impacts 2026 program. The research focuses on the compression and efficient processing of graph data, bridging the fields of graph signal processing, graph machine learning, and geometric deep learning. The project aims to develop principled methods for graph and signal compression, with theoretical guarantees for recovery and downstream learning, enabling efficient and trustworthy learning on large-scale graph data. The research will combine solid mathematical foundations, including sampling, reconstruction, and spectral theory, with modern graph neural networks (GNNs) and higher-order graph learning. The work will be conducted in close collaboration with the University of Southern California, offering an international research environment. Applicants should have or be completing an M2 or final-year engineering degree in applied mathematics, signal processing, computer science, or related fields. Strong foundations in signal processing and machine learning, proficiency in Python and PyTorch, and a keen interest in theory-driven research in graph signal processing and geometric deep learning are required. The position is based at Télécom Paris, Palaiseau, with a start date in September or October 2026 and a duration of 3 years. Funding is provided by the IMT Futur, Ruptures & Impacts 2026 program, covering stipend and tuition. The supervisory team includes Jhony H. Giraldo (Télécom Paris), Aref Einizade (Télécom SudParis), and Antonio Ortega (University of Southern California). Research topics include graph signal processing, graph neural networks, graph coarsening, spectral theory, and applications to large-scale graph-structured data such as social networks, recommender systems, and biological networks. The project will address challenges in compressing graph signals and topology, developing frameworks with decoding and task-level guarantees, and leveraging higher-order structures like cliques and motifs for efficient learning. Applications must be submitted via the official IMT PhD platform. The deadline is February 15, 2026. Incomplete applications will not be considered. For more information, refer to the full PhD description and contact the supervisors as needed.