Karteek Alahari

Karteek Alahari is a Researcher (Directeur de Recherche) at Inria Grenoble - Rhône-Alpes, leading the Thoth project-team. He also holds associate memberships with the Visual Geometry Group at the University of Oxford and the WILLOW team at École Normale Supérieure (ENS). His research focuses on visual understanding with large-scale datasets, particularly learning robust visual representations under partial supervision, spanning frameworks like incremental learning, weakly-supervised learning, and adversarial training.

Research Focus

His work addresses challenges in computer vision, including semi-supervised instance segmentation, unsupervised video domain adaptation, and self-supervised contrastive representation learning. He explores intersections between deep learning and graphical models, emphasizing spatio-temporal relations in videos and geometric models for object recognition.

Scientific Leadership & Recognition

• Deputy Scientific Director in charge of AI at Inria (2024)
• Co-Director of the PEPR IA research programme (2024)
• Area Chair for CVPR 2025, ECCV 2024, ICCV 2023, and others
• Editorial board member of IJCV (2019–2023) and CVIU (2018–2023)

Students & Collaborations

He has co-advised numerous PhD students, including Jules Bourcier (2021–2024), Lina Mezghani (2019–2023), and Zhiqi Kang (2021–present). Former advisees like Anand Mishra and Yang Hua are now faculty members at IIT Jodhpur and Queen’s University Belfast, respectively.