Alaa Mazouz

Alaa Eddine Mazouz

Associate Professor in Embedded AI — Télécom Paris

Maître de Conférences, équipe SSH — Télécom Paris

Associate Professor, SSH team — Télécom Paris

Embedded AI Frugal Deep Learning FPGAs & Reconfigurable HW Continual Learning Computer Vision

Présentation

About

Alaa Mazouz a rejoint l’équipe SSH de Télécom Paris le 1er septembre 2025 en tant que Maître de Conférences. Ses recherches portent sur l’intelligence artificielle embarquée, le deep learning frugal et l’optimisation de modèles pour des systèmes contraints en ressources, en particulier sur FPGA et architectures reconfigurables. Il s’intéresse aussi à l’apprentissage en continu, à l’adaptation en ligne et aux applications de vision par ordinateur et d’IA embarquée pour les systèmes spatiaux.

Il a obtenu son doctorat à l’Université de Surrey (UK), au Surrey Space Centre, sur le deep learning et la vision pour satellites (projet UKSA & ESA). Il a ensuite réalisé un postdoctorat à Télécom Paris (SSH et C2S) sur la compression d’images par réseaux de neurones, l’IA embarquée et l’apprentissage adaptatif sur FPGA.

Alaa Mazouz joined the SSH team at Télécom Paris on September 1st 2025 as an Associate Professor in Embedded AI. His research focuses on embedded AI, frugal deep learning, and model optimization for resource-constrained systems, particularly on FPGAs and reconfigurable hardware. He is also interested in continual learning, online adaptation, and computer-vision applications for space systems.

He received his PhD from the University of Surrey (UK) at the Surrey Space Centre (UKSA and ESA-funded project) . He then completed a postdoctoral fellowship at Télécom Paris (SSH and C2S), focusing on neural image compression, embedded AI, and adaptive learning on FPGA.

Thèmes de recherche

Research Interests

  • Embedded AI on FPGAs • Reconfigurable & runtime-adaptive architectures
  • Frugal / efficient deep learning under tight compute & memory budgets
  • Continual & online learning; on-device training and experience replay
  • Neural image/video compression; watermarking and secure deployment
  • Computer vision for space systems; on-board perception

Publications sélectionnées

Selected Publications

  • [Preprint] Lightweight Embedded FPGA Deployment of Learned Image Compression with Knowledge Distillation and Hybrid Quantization, A. Mazouz et al. — arXiv:2503.04832 (cs.CV), IEEE TCSVT 2025. arXiv
  • [Preprint] Security and Real-time FPGA Integration for Learned Image Compression, A. Mazouz et al. — arXiv:2503.04867 (cs.CR), IEEE TMM 2025. arXiv
  • [Preprint] An FPGA Compiler for On-the-Fly Adaptive CNN Deployment and Reconfiguration, A. Mazouz, D. H. Le & V-T. Nguyen — arXiv:2504.08534 (cs.AR), IEEE TCAD 2025. arXiv
  • Mazouz, A. & Nguyen, V.-T. Online Continual Streaming Learning for Embedded Space Applications, Journal of Real-Time Image Processing, 2024. DOI
  • Mazouz, A. & Nguyen, V.-T. Automated Runtime Reconfiguration of CNNs for Embedded AI, Journal of Real-Time Image Processing, 2024. DOI
  • Mazouz, A. Online Reconfiguration of CNNs for Onboard Vision Applications, PhD Thesis, University of Surrey, 2022. DOI
  • Mazouz, A. & Bridges, C. P. Automated CNN back-propagation pipeline generation for FPGA online training, Journal of Real-Time Image Processing, vol. 18(6), pp. 2583–2599, 2021. DOI
  • Mazouz, A. & Bridges, C. P. Automated Offline DSE and Online Reconfiguration for CNNs, IEEE EAIS, 2020; Journal of Evolving Systems, 2022. DOI
  • Mazouz, A. & Bridges, C. P. Adaptive Hardware Reconfiguration for Performance Trade-offs in CNNs, NASA/ESA AHS, 2019. IEEE Xplore
  • Mazouz, A. & Bridges, C. P. Multi-Sensory CNN Models for Close Proximity Satellite Operations, IEEE Aerospace Conference, 2019. IEEE Xplore

Full list: ResearchGate