PhD in Machine Learning: Model Debiasing, Foundation Models, and Efficient Deep Learning at University of Turin University of Turin in Italy

PhD candidates are invited for a funded position at the University of Turin in machine learning, with research focused on three closely connected themes: model debiasing, foundation models, and efficient deep learning. The project asks questions such as how to measure and reduce unwanted bias in large models, how pretrained foundation models learn and fail, and how to make models faster, lighter, and cheaper without losing capability.

The position is framed in the context of an Italian National Project and includes access to substantial compute resources, including an internal GPU cluster at the University of Turin and the possibility of additional computation on the Leonardo national cluster facility. The poster also mentions a possible co-tutelle arrangement with Télécom Paris (Institut Polytechnique de Paris), which could lead to a double PhD degree, although this is still to be confirmed.

Ideal candidates should have strong ML fundamentals, solid programming ability in PyTorch or JAX, and the ability to work quickly and independently: reading papers, reimplementing ideas, and running experiments in days. The call emphasizes rigor in evaluation and honesty about negative results.

The official call is call n.45 in the Computer Science section, titled Selezione di architetture neurali robuste a polarizzazione e pregiudizi. The application deadline is 9 June 2026 at 11:59 a.m. Italian time. Applicants should use the linked application portal and are encouraged to apply well before the deadline.