🦕 Donald Shenaj

I am a Postdoctoral Researcher at the University of Pisa, working in Davide Bacciu’s group. I am part of the CoEvolution Project for safe and trusthworthy AI.

I obtained my Ph.D. at the University of Padova, under the guidance of Prof. Pietro Zanuttigh. Before that, I earned a Master Degree from the same University, and a Bachelor Degree from the University of Bologna, both with honors.

My academic journey has been marked by collaborations with globally renowned institutions. From October 2024 to April 2025 I as a research intern at Samsung Research UK under the supervision of Umberto Michieli and Mete Ozay. From May 2023 to November 2023, I was a research intern at Mila - Quebec AI Institute & Concordia University, working with Prof. Eugene Belilovsky.

During my Ph.D., I addressed fundamental limitations of machine learning models for visual understanding, specifically their challenges in adapting to dynamic, real-world conditions, learning continuosly over time and from decentralized data sources, My research explored these challenges through three emerging paradigms: domain adaptation, continual learning, and federated learning.

Building on this foundation of adaptability and decentralized learning, my current work focuses on personalized generative AI, where I explore how model merging, image generation, and large language models (LLMs) can be combined to enable personalized and adaptive AI systems.

I am always open to new collaborations! Email me, if you would like to discuss or catch up at a conference.

[Download CV]

Reviewer

CVPR 2026, ICCV 2025, CVPR 2025 (Outstanding Reviewer), AAAI 2025, ICLR Workshops 2025, CVPR Workshops 2025, WACV 2025, IEEE TPAMI, IEEE TMM, Pattern Recognition, CVIU, MMSP 2024, ICML 2023 Workshops, ICPR 2022, Harms and Risks of AI in the Military (HRAIM)

Awards

Best Paper Award @ ICCV Workshop on Personalization in Generative AI
Outstanding Reviewer @ CVPR 2025
National PhD scholarship on a free subject
M.Sc., Summa Cum Laude, University of Padova
B.Sc., Summa Cum Laude, University of Bologna
Regional Grant for Academic Excellence in High School, Emilia-Romagna Region
High School Award, project presented at the Italian selection for EUCYS and Mostratec

Publications

Preprints

	K-Merge: Online Continual Merging of Adapters for On-device Large Language Models D. Shenaj, O. Bohdal, T. Ceritli, M. Ozay, P. Zanuttigh, U. Michieli under review, 2025. [paper]
	FedPromo: Federated Lightweight Proxy Models at the Edge Bring New Domains to Foundation Models M. Caligiuri, F. Barbato, D. Shenaj, U. Michieli, P. Zanuttigh under review, 2025. [paper][code]

Conferences

	LoRA.rar: Learning to Merge LoRAs via Hypernetworks for Subject-Style Conditioned Image Generation D. Shenaj, O. Bohdal, M. Ozay, P. Zanuttigh, U. Michieli IEEE/CVF International Conference on Computer Vision (ICCV), 2025. [paper][website][code][video]
	Adaptive Local Training in Federated Learning D. Shenaj, E. Belilovsky, P. Zanuttigh International Conference on Learning Representations (ICLR), 2025, MCDC Workshop, and EUSIPCO 2025. [paper][code]
	When Cars meet Drones: Hyperbolic Federated Learning for Source-Free Domain Adaptation in Adverse Weather G. Rizzoli, M. Caligiuri, D. Shenaj, F. Barbato, P. Zanuttigh IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2025 (oral). [paper][code]
	Source-Free Domain Adaptation for RGB-D Semantic Segmentation with Vision Transformers G. Rizzoli, D. Shenaj, P Zanuttigh IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), Pretrain Workshop 2024 (oral). [paper]
	Asynchronous Federated Continual Learning D. Shenaj, M. Toldo, A. Rigon, P. Zanuttigh IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), FedVision Workshop, 2023 (oral). [paper][code]
	Learning Across Domains and Devices: Style-Driven Source-Free Domain Adaptation in Clustered Federated Learning D. Shenaj, E. Fanì, M. Toldo, D. Caldarola, A. Tavera, U. Michieli^†, M. Ciccone^†, P. Zanuttigh^†, B. Caputo^† IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2023 (oral). [paper][code]

Journals

	Federated Learning in Computer Vision D. Shenaj, G. Rizzoli, P. Zanuttigh IEEE Access, 2023. [paper]
	Continual coarse-to-fine domain adaptation in semantic segmentation D. Shenaj, F. Barbato, U. Michieli, P. Zanuttigh Elsevier Image and Vision Computing, 2022. [paper][website][code]

^* Equal contribution. ^† Equal supervision.

Ph.D., Master’s and Bachelor’s Thesis

[T3] D. Shenaj, “Morphing Distributed and Transfer Learning Paradigms for Visual Understanding”, Ph.D thesis in Information Engineering, University of Padova, March 2025.

[T2] D. Shenaj, “Coarse-to-Fine Learning for Semantic Segmentaion across Multiple Domains”, MS thesis in ICT for Internet and Multimedia, University of Padova, September 2021.

[T1] D. Shenaj, “Implementation and analysis of a vehicle counter system with Python and OpenCV”, BS thesis in Electronics Engineering for Energy and Information, University of Bologna, October 2019.