Francesco Tudisco
Reader in Machine Learning
- School of Mathematics
- Maxwell Institute for Mathematical Sciences
- GAIL (Generative AI Laboratory)
Contact details
- Email: f.tudisco@ed.ac.uk
Address
- Street
-
James Clerk Maxwell Building, Peter Guthrie Tait Road
- City
- Edinburgh
- Post code
- EH9 3FD
Background
Francesco Tudisco is a Reader (Associate Professor) in Machine Learning in the School of Mathematics at the University of Edinburgh.
Before joining Edinburgh in September 2023, he was Assistant Professor in Applied Mathematics at the Gran Sasso Science Institute (GSSI), Italy, where he continues to hold a part-time Associate Professorship. Previously, he was a Marie Skłodowska-Curie Individual Fellow at the University of Strathclyde.
His research lies at the intersection of mathematics, machine learning and scientific computing, with a particular focus on developing mathematical foundations and scalable algorithms for modern AI and its applications to scientific discovery. He is also actively involved in translating research into practice through collaborations with industry and entrepreneurship.
Research summary
Francesco’s research develops mathematical methods and scalable algorithms for modern artificial intelligence, with a particular emphasis on making machine-learning systems more efficient, reliable and capable of addressing challenging scientific problems.
A major strand of his research concerns efficient and geometry-aware machine learning. This includes low-rank and tensor-structured neural networks, parameter-efficient fine-tuning of foundation models, Riemannian and constrained optimisation, and methods that reduce the computational, memory and energy requirements of training large models.
A second strand focuses on AI for science and scientific machine learning, including neural solvers and generative models for physical systems, partial differential equations and scientific simulation. His work investigates how mathematical structure, physical knowledge and dynamical-systems principles can be incorporated into machine-learning architectures to improve their efficiency, generalisation and interpretability.
His broader interests include graph neural networks and higher-order network models, nonlinear spectral methods, dynamical systems, numerical linear algebra and optimisation. A recurring theme across these areas is the use of mathematical structure to design machine-learning methods that are both theoretically principled and practically scalable.
Knowledge exchange
His research has led to substantial collaborations with industry. These include leading an industrial research programme with Leonardo SpA on quantum machine learning and combinatorial optimisation, alongside projects and partnerships applying advanced AI and optimisation methods to scientific and industrial problems.
His research group also develops methods whose effectiveness has been demonstrated through international challenges. In 2025, work on neural oscillatory architectures developed by the group led to first place in the ACM ICAIF Cryptocurrency Forecasting Competition, outperforming approximately 130 teams and 740 submitted models.
Francesco was selected as a Bayes Innovation Fellow in 2024, supporting the translation of research in efficient machine learning into commercial applications. More broadly, his knowledge-exchange activities aim to connect mathematical innovation, frontier AI research and practical deployment, particularly in areas where efficiency, reliability, scientific structure and trustworthy decision-making are important.