Postdoctoral Researcher in Data-Efficient Visual Learning at Leiden University

We are recruiting a postdoctoral researcher to join the Machine Learning cluster at LIACS, Leiden University, working closely with Dr. Hazel Doughty on adaptive visual learning.

The research will address a key limitation of current visual foundation models: although they perform well on broad, general-purpose tasks, they often struggle to acquire the specialised knowledge and fine-grained distinctions needed in real-world domains, especially when labelled data is scarce. The project aims to develop models that can specialise efficiently by exploiting structure within their representations and reusing knowledge across tasks.

Potential directions include foundation model adaptation, fine-grained visual understanding, self-supervised and data-efficient learning, transfer across tasks, structured and compositional learning, and vision-language or multimodal models. The successful candidate will have considerable freedom to shape the research agenda and develop ambitious new ideas.