Alfie McGlennon

I’m finishing an MSc in Climate Change and AI at the University of Reading, after a BSc in Meteorology and Climate in the same department. I got hooked on the gap between what climate models simulate and what actually happens on the ground, particularly for compound events where multiple hazards arrive together. My dissertation focuses on compound heat stress across Europe: when heat and humidity combine in ways that neither variable alone would predict. I’m trying to work out where climate models get the dependence wrong, and whether machine learning can fix it directly.

I was drawn to reinforcement learning through the visually compelling demos of agents learning to drive cars or play games. I wanted to test it on something I understood. Weather and climate systems do not map cleanly to the typical RL framing, but energy grids, where weather is a core input, were a much better fit. So I started building an RL environment for GB electricity dispatch. The scope kept growing. What began as ‘can an agent learn merit order?’ turned into a serious exercise in data assimilation from public sources, and a test of how far you can get with human-led architecture and AI execution. The GB Grid Scenario Tool ended up as a validated DC power flow model of the transmission network, and the RL agents trained on top of it went from spending £30 billion a year to within 6% of real grid costs. Alongside that, the copula explainer I wrote for my own understanding turned into an interactive visual guide. I get hooked on problems and follow them further than the brief asks for.

I’m targeting roles in climate risk analytics, reinsurance (Swiss Re and Munich Re type work), or operational meteorology (ECMWF-adjacent). Based in the Reading area. The through-line across all my work is applying statistical and machine-learning methods to climate and weather data where the outputs have operational or commercial value.