Biography

Researcher turned software engineer. I recently completed a PhD in Computer Science with Dr He Sun at Edinburgh University.

My research focused on the Algorithmic Kadison-Singer problem, graph clustering, and connectomics. I was awarded a BSc in Computer Science and Mathematics (first class) and an MSc in Data Science (distinction) from the University of Edinburgh. Along with my research, I am interested in real-time entity resolution and its applications.

I love exploring the Highlands by foot, bike, and kayak. I'm also a keen Fives player and ran the Edinburgh based Caledonian Fives team.

Publications

• Fully Dynamic Coreset Spectral Clustering (ICML 2026), with Gregory Schwartzmann and Peter Macgregor. We extend Coreset Spectral Clustering to the fully dynamic graph setting and beat the best existing dynamic spectral clustering algorithms on query time and update time, in theory and practice.
• Distributed control circuits across a brain-and-cord connectome (Nature 2026): with many others 2025. A massive collaboration to produce the first connectome of a fly brain and nerve cord. NTAC was used to successfully infer the types of neurons in a partially constructed connectome.
• Connectivity Is All You Need: Inferring Neuronal Types with NTAC (Nature Communications 2026, Spotlight): with Gregory Schwarztman, David García-Soriano & Arie Matsliah. We give a pair of novel semi-supervised and unsupervised algorithms for inferring the types of neurons in the full fly brain connectome, using the connectivity of the neurons.
• Mini-Batch Kernel $k$-means (arXiv): with Gregory Schwarztman 2024. We give the first mini-batch algorithm for kernel k-means with sublinear iteration time.
• Coreset Spectral Clustering (ICLR 2025, talk): with Gregory Schwarztman, Peter Macgregor & He Sun 2025. We give the first coreset algorithm for spectral clustering that allows spectral clustering to scale to gigantic graph datasets.
• Polynomial-Time Algorithms for Weaver's Discrepancy Problem in a Dense Regime (arXiv): with Peter Macgregor & He Sun 2024. We give a polynomial time algorithm for Weaver's Discrepancy Problem when the number of vectors is at least quadratic in the dimension.
• Is the Algorithmic Kadison-Singer Problem Hard? (ISAAC 23): with Peter Macgregor & He Sun 2022. We show that the Algorithmic Kadison-Singer problem is FNP-hard in certain regimes and give the first quasi-polynomial time algorithm, in the number of vectors.
• On the spectral method for finding shortest paths (MSc dissertation (distinction)): Characterising the solution to an open question on the hot spots conjecture on graphs. 2021.

Experience

• Summer 2023-: Part time Data Science contractor for Police Scotland, working on detecting organised crime selling stolen bikes online.
• Summer 2019: Data Science Intern at Featurespace, worked on a feature extraction pipeline for application fraud.

Education

• 2021 - Present: PhD Student, University of Edinburgh
• 2020 - 2021: MSc Data Science, University of Edinburgh
• 2016 - 2020: BSc Computer Science and Mathematics, University of Edinburgh