Machine Learning × Energy Systems
I'm an ML researcher and engineer working at the intersection of machine learning and energy systems. I'm currently based in Berlin at the Merantix AI Campus Hacker Room, building open-source tools for embedding visualization and representation learning.
Previously, I was an ML Scientist at GE Vernova (2022–2025), where I was the sole ML engineer on the Constellation project (£17.8M, UK Power Networks) — developing power flow prediction and anomaly detection models for smart substations. Before that, I was a first engineer at PyPSA meets Earth, contributing to the world's first open-source global energy system model.
I hold a BSc (Hons) in Artificial Intelligence and Mathematics from the University of Edinburgh. My research interests span time-series forecasting for power grids, representation learning, AI safety, and computer vision for energy infrastructure.
Publication list aligned with Google Scholar (117 citations).
PyPSA-Earth. A New Global Open Energy System Optimization Model Demonstrated in Africa
Applied Energy 341, 121096 — 2023
PyPSA meets Africa: Developing an Open Source Electricity Network Model of the African Continent
2021 IEEE AFRICON, 1–6 — 2021
IET Conference Proceedings CP922 2025 (14), 2707–2711 — 2025
Open-source embedding visualization toolkit with Euclidean and non-Euclidean views for analyzing learned representations. CLIP-based image embeddings and HuggingFace Spaces deployment. [demo] [code]
Python package for global power infrastructure data extraction from OpenStreetMap. Used by the Polish TSO and energy non-profits. Presented at Stanford and Princeton. [PyPI] [code] [talk]
Interactive knowledge graph of 5000+ AI alignment papers using LLM-based clustering and graph visualization for literature exploration. [site] [code] [blog]
Core contributor to the first open-source global energy system optimization model, including major performance improvements to network build pipelines. [code] [docs]
Satellite imagery detection of power infrastructure (transmission towers) using R-CNN and CycleGANs for domain adaptation across landcover types. [code]
Unified web and code search MCP server with weighted provider routing and automatic failover across multiple search backends. [code]
Merantix AI Campus, Berlin — Hacker Room Residency
GE Vernova — Remote (Edinburgh / London / Berlin)
PyPSA meets Earth — Remote (Edinburgh)
University of Edinburgh, School of Informatics