Postdoctoral Position, Geometric Deep Learning at Télécom Paris | European Laboratory for Learning and Intelligent Systems

We are opening a postdoctoral position at Télécom Paris (Institut Polytechnique de Paris) within the LTCI laboratory, in the area of geometric deep learning and graph machine learning.

This postdoctoral position is part of a broader research program focused on understanding why and when geometric inductive biases improve learning, and on designing continuous and higher-order models with theoretical guarantees. The project builds on recent work on continuous product graph neural networks and continuous simplicial neural networks, with the goal of advancing principled and scalable learning methods for structured data.

The research lies at the intersection of:

Geometric and topological deep learning
Graph signal processing
Continuous and PDE-inspired learning models
Stability, expressivity, and generalization analysis of geometric deep learning models

The postdoc will work under the primary supervision of Jhony H. Giraldo, with opportunities for scientific interaction and collaboration with:

Antonio Ortega (University of Southern California)
Fragkiskos Malliaros (CentraleSupélec)
Dorina Thanou (EPFL)

Candidate profile (indicative):

PhD (or near completion) in applied mathematics, computer science, electrical and computer engineering, or related fields.
Strong background in machine learning, with interest in theoretical aspects.
Experience in geometric deep learning and/or graph signal processing, including graph neural networks and/or higher-order neural network models.
A strong publication record in machine learning, graph signal processing, or related fields is expected.
Good programming skills in Python / PyTorch
Ability to work independently and propose original research directions

Practical information:

Location: Télécom Paris, Palaiseau (France).
Start date: Spring / Summer 2026 (May–June).
Duration: 1 year, renewable.
Funding: ANR, the French National Research Agency JCJC project DeSNAP.

Applications are reviewed on a rolling basis.