Maysam Behmanesh

Linkedin , Google Scholar , GitHub

About me

I am an AI Research Scientist at the BioMed-X Research Institute, Paris-Saclay, France, focusing on generative AI and deep diffusion models for computational structural biology and bispecific antibody design. My research interests include AI methods for understanding data through geometry, structure, and multimodality, particularly geometric deep learning and multimodal machine learning for graphs and complex data.

Previously, I was a Postdoctoral Researcher in the GeomeriX group at the LIX research laboratory of École Polytechnique IP-Paris, from April 2022 to January 2025, working with Prof. Maks Ovsjanikov. My research focused on geometric deep learning and multimodal machine learning, with an emphasis on graphs and multimodal data.

I earned my PhD in Computer Engineering-Artificial Intelligence from the University of Isfahan, where I was a member of the ILS-Lab and worked with Prof. Peyman Adibi. During my PhD, I was also a visiting researcher at GIPSA-Lab at Grenoble INP (2019–2020), working with Prof. Jocelyn Chanussot.

Earlier in my research career, I worked on chaotic time-series prediction, imbalanced classification, neuro-fuzzy systems, and evolutionary computation.

News

• Two papers accepted at ICML 2026:
  • "Beyond ReLU: Bifurcation, Oversmoothing, and Topological Priors" (Spotlight), with Erkan Turan, Gaspard Abel, Emery Pierson, and Maks Ovsjanikov.
  • "Graph Alignment via Dual-Pass Spectral Encoding and Latent Space Communication", with Erkan Turan and Maks Ovsjanikov.

• I gave a talk about my recent papers in geometric deep learning at SAMOVAR - Télécom SudParis, Institut Polytechnique de Paris

• I gave a talk at Inria Paris, hosted by the Argo research team. My talk was on "Enhancing Graph Neural Networks with Geometric Structure Analysis".
• Our new paper, Smoothed Graph Contrastive Learning via Seamless Proximity Integration, with Maks Ovsjanikov has been accepted at Learning on Graphs Conference LoG-2024 PDF.

• Our new paper, Cross-Modal and Multimodal Data Analysis Based on Functional Mapping of Spectral Descriptors and Manifold Regularization, with Peyman Adibi, Jocelyn Chanussot and Sayyed Mohammad Saeed Ehsani is published in Neurocomputing PDF.

• I gave a talk at Télécom Paris, hosted by the S2A team. My talk was on "Graph Representation Learning for Multimodal Data - Challenges and Innovative Methods".

• Our paper "TIDE: Time Derivative Diffusion for Deep Learning on Graphs", with Maximilian Krahn and Maks Ovsjanikov has been accepted at ICML 2023, PDF.