Hi, I’m Mateo.

I build ML systems

I hold a PhD in Network Science from UCL and the Alan Turing Institute, where I studied how graph structure and information propagation shape human behaviour. My research sits at the intersection of graph theory, statistical modelling, and large-scale computation.

I’m currently a Senior Machine Learning Engineer at Vu.City, building ML systems for 3D city-scale reconstruction and semantic understanding. Previously, I spent 8 years at Foster + Partners as an Associate Partner, leading data science and deploying ML pipelines that operated at an urban scale. Before that, I developed transport optimisation algorithms at SignalBox.

Across research and industry, my work has centred on taking mathematically grounded ideas and turning them into production systems that function under real-world constraints: noisy data, long feedback loops, and high-stakes decisions. I operate end-to-end: from formulation and prototyping through to engineering, deployment, and stakeholder alignment.

I’m interested in problems that demand both mathematical depth and engineering rigour:

  • graph-structured and high-dimensional data
  • probabilistic modelling and inference
  • systems where model decisions carry real consequences

WRITING

The Alpha-Blending Problem: Semantic Segmentation in 3D Gaussian Splatting

A survey of why making 3D Gaussian Splatting semantically meaningful is harder than it looks. Covers the core technical challenge — alpha-blending ambiguity at object boundaries — and the three paradigms that have emerged across the CVPR, ECCV, and ICCV 2024–25 literature: feature-field distillation, 2D-to-3D lifting, and identity-centric methods.

Spatial Interaction Models

An in-depth exploration of spatial interaction models with interactive visualizations, covering their theoretical foundations, mathematical formulation, and practical applications in urban planning.

Transport Networks

Interactive exploration of centrality measures of the metro systems of London, New York, Chicago, and Santiago de Chile. Time-weighted graphs constructed from publicly available GTFS data.