CV
General Information
| Full Name | Souhail Hadgi |
| hadgisouhail@gmail.com | |
| Phone | +33 7 67 70 58 62 |
| linkedin.com/in/shadgi |
Summary
- Final-year PhD student specializing in Artificial Intelligence and Computer Vision. My research focuses on transfer learning for 3D data, with emphasis on effective fine-tuning of pre-trained models for shape analysis, segmentation, and cross-modal (image and text) applications.
Technical Skills
| Languages | Python, SQL, MATLAB, R |
| Frameworks | PyTorch, TensorFlow, Scikit-learn |
| Areas of Expertise | Machine Learning, Deep Learning, 2D & 3D Computer Vision, Natural Language Processing, LLM, Generative Modeling, Computer Graphics, Optimization |
Education
-
Jan 2023 – 2026 PhD Student
École Polytechnique, France - Supervised by Prof. Maks Ovsjanikov.
-
2022 M2 Master: MVA (Mathematics, Vision, Learning)
École Normale Supérieure Paris-Saclay, France - {"GPA"=>4.0}
-
2022 Master of Engineering (Major in Mathematics, Specialization in Data Science)
CentraleSupélec, France - {"GPA"=>4.0}
-
2019 Bachelor of Engineering
CentraleSupélec, France
Publications
-
2025 PatchAlign3D: Local Feature Alignment for Dense 3D Shape Understanding
ArXiv Preprint - Souhail Hadgi, Bingchen Gong, Ramana Sundararaman, Emery Pierson, Lei Li, Peter Wonka, Maks Ovsjanikov.
-
2025 Escaping Plato’s Cave: Towards the Alignment of 3D and Text Latent Spaces
CVPR 2025 - Souhail Hadgi, Luca Moschella, Andrea Santilli, Diego Gomez, Qixing Huang, Emanuele Rodolà, Simone Melzi, Maks Ovsjanikov.
-
2024 To Supervise or Not to Supervise: Understanding and Addressing the Key Challenges of Point Cloud Transfer Learning
ECCV 2024 - Souhail Hadgi, Lei Li, Maks Ovsjanikov.
Experience
-
May 2022 – Oct 2022 Research Intern
École Polytechnique - Analysis of several unsupervised pre-training approaches for 3D representation learning.
- Adapted scene point-level contrastive learning approaches for 3D shapes.
-
Aug 2020 – Jul 2021 Data Scientist Intern
DataScientest - Designed Computer Vision modules for a Deep Learning curriculum.
- Supervised Deep Learning applied projects.
- Conceived the scientific content of data challenges.
- Taught Deep Learning courses for cohorts of learners.
Projects
-
Reconcile video predictions from multiple angles
Sicara - Created a TensorFlow pipeline that generates videos of highways corresponding to a different view angle from the initial viewpoint.
- Optimized an auto-encoder architecture for image generation.
-
Segmentation models for audio data
Illuin Technology - Created an end-to-end PyTorch deep learning pipeline for speaker diarisation.
- Trained and evaluated the diarisation pipeline on the TCOF dataset.