Research
My research studies how relational structure can become a useful inductive bias. Graphs represent pairwise interactions; hypergraphs and simplicial or cellular complexes represent higher-order interactions. The goal is to build models that use this structure without sacrificing scalability, stability, or interpretability.
Active funded programs
DeSNAP — Deep Simplicial Neural Networks for Advanced Geometry Processing
ANR · Principal Investigator · 2025–2029 · €337,862
DeSNAP studies neural networks for higher-order geometric structures, with emphasis on continuous formulations, theoretical properties, and scalable learning.
SIGMA — Signal-aware Graph Summarization with GNN Guarantees
Hi! PARIS Synergy Fellowship · Co-Principal Investigator · 2026–2029 · €200,000
SIGMA connects graph summarization with signal preservation and guarantees for downstream graph-neural-network behavior.
Compressing Graph Data: From Signal Processing to Machine Learning
IMT Futur, Ruptures & Impacts · Supervisor · 2026–2029 · €155,000
This doctoral project develops principled graph compression methods across signal-processing and machine-learning perspectives.
SODA — System on Chip Design Leveraging Artificial Intelligence
ANR · Team member · 2023–2027 · €531,000 consortium funding
SODA explores how artificial intelligence can support circuit and system-on-chip design.
Translation and collaboration
My program includes collaboration with academic and industrial partners, including TotalEnergies and Idemia. Research outputs include open scientific publications as well as patent applications.
Collaboration
I welcome research discussions that have a clear connection to graph machine learning, geometric deep learning, graph signal processing, or application areas I’m interested in. A short message describing the scientific question, relevant background, and intended form of collaboration is particularly helpful.