My current research sits at the intersection of machine learning, statistics, and AI security. I am particularly interested in mathematically grounded methods for making modern machine-learning systems both more trustworthy and more scalable. One strand of my work studies privacy and security risks in machine learning, especially training-data leakage in transfer learning, together with practical defenses based on privacy-preserving computation. This includes developing and evaluating reconstruction attacks, studying their robustness under realistic threat models, and designing defenses that protect sensitive training information while preserving utility.

A second strand of my research focuses on highly scalable probabilistic modelling, especially Gaussian process regression for massive datasets. I work on nearest-neighbour-based Gaussian process methods that retain the principled uncertainty quantification of Gaussian processes while scaling to modern large-data settings, and on the statistical theory that explains their consistency, optimal convergence rates, calibration, and robustness to hyperparameter choice. Across both areas, my aim is to combine rigorous mathematical analysis with practical algorithms and large-scale computational experiments.

Past research interests:

• Quantum information and quantum computing

• Mathematical physics and quantum many-body systems

• Topology of configuration spaces, braid groups, and quantum statistics on graphs

• Anyons and non-Abelian braiding on networks

• Reduced density matrix functional theory and generalized Pauli constraints

• Inverse problems, matrix reconstruction, and spectral methods