Computational science · LLM research · Energy systems

EllyessBenmoufok

PhD computational scientist at Imperial College London. The throughline across my work is scientific and computational methods validated against real-world ground truth. In energy, that means bias-correcting ERA5 reanalysis against observed wind generation (PyVWF) and feeding it into continental-scale energy system optimisation (PyPSA-Eur). In LLM research, it means scoring populations of model personas against nationally representative survey distributions. I build modular scientific Python systems with an emphasis on reproducibility, performance, and clear handoff between simulation and decision models.

Get in touchView research
01

Education

Oct 2021 — 2026

Department of Earth Science and Engineering

PhD, Computational Science

Imperial College London

  • Developed PyVWF — modular wind simulation and reanalysis bias-correction tooling.
  • Built scalable ERA5 spatiotemporal pipelines integrated into PyPSA-Eur optimisation studies.
  • Published peer-reviewed work on high-resolution wind bias correction.

Oct 2019 — Oct 2020

Distinction

MSc, Applied Computational Science and Engineering

Imperial College London

  • Numerical methods, optimisation, and parallel programming in Python and C++.
  • Research project on multi-output regression with GANs (co-authored publication).

Sep 2015 — Jun 2018

First Class Honours

BSc, Physics

University of Surrey

  • Foundations in mathematical modelling, numerical analysis, and physical systems.
  • Applied computational techniques (Fortran, Python) to scientific problem-solving.
02

Experience

Oct 2020 — Oct 2024

Graduate Teaching Assistant

Imperial College London

  • Supported teaching and assessment across computational science modules.

Jun 2021 — Sep 2021

Data Science Intern

Shell

  • Applied data science methods to energy-relevant datasets; delivered analysis and prototypes.

Dec 2020 — Mar 2021

App Developer

Imperial College London

  • Built geospatial metrics and ML-driven spatial analyses on Google Earth Engine.

Jan 2019 — Aug 2019

Data Scientist

Link Humans

  • Most technical member of the team; led automation of the Employer Brand Index, the company's core data product.
  • Built web scrapers and data-cleaning pipelines to source and standardise employer branding signals.
  • Prototyped the data pipeline intended to support training a language model on the collected corpus.
03

Selected Publications

Energy Conversion & Management · 2026

Geographic variability in reanalysis wind speed biases: A high-resolution bias correction approach for UK wind energy

Wang, Y., Warder, S., Benmoufok, E.F., Wynn, A., Buxton, O.R.H., Staffell, I., & Piggott, M.D.

Energy Conversion and Management, 352, 121066.

Extends the multi-country high-resolution bias correction framework PyVWF.

View publication

Energy · 2024

Improving wind power modelling through granular spatial and temporal bias correction of reanalysis data

Benmoufok, E.F., Warder, S., Zhu, E., Bhaskaran, B., Staffell, I., & Piggott, M.D.

Energy, 313, 133759.

Lead-author study introducing PyVWF — a multi-country high-resolution bias correction framework for reanalysis-driven wind power modelling.

View publication

Applied Sciences · 2022

Multi-Output Regression with Generative Adversarial Networks (MOR-GANs)

Phillips, T.R.F., Heaney, C.E., Benmoufok, E., Li, Q., Hua, L., Porter, A.E., Chung, K.F., & Pain, C.C.

Applied Sciences, 12(18), 9209.

Co-authored study developing GAN approaches for multi-output regression in scientific modelling.

View publication
04

Technical Stack

Languages

Python · JavaScript · C++ · Fortran · GDScript

Scientific

NumPy · Pandas · Xarray · SciPy · Dask

Energy / Optimisation

PyPSA · PyPSA-Eur · Atlite · Gurobi

Geo / Climate

GeoPandas · Shapely · Rasterio · ERA5

Workflow

Snakemake · Conda · Git · GitHub

Writing

LaTeX · Matplotlib

05

Beyond Research

Outside computational energy modelling, I work on real-time audio-visual systems and creative coding — generative visuals in TouchDesigner and music production in Ableton Live, oriented toward interactive and performance-driven workflows.

I'm also developing Fool's Ascension — a roguelite card game built on Russian Durak. Branching 3-act runs, persistent meta-progression, and tightly-systemised encounter design in Godot.

06

Selected Work

01LLM Research

LLM Personas vs Real UK Opinion

An independent study of whether populations of LLM personas reproduce the real distribution of opinions a human group holds, measured against nationally representative UK surveys. Five persona methods, two models, distribution-fidelity metrics (TVD/JSD), and diagnostics that detect a model being right for the wrong reasons.

Open
02Research Engineering

PyVWF: Python Virtual Wind Farm

Methods for reducing systematic error in wind resource estimation pipelines, with emphasis on robust calibration and reproducible evaluation.

Open
03Energy Systems

Improving Wind Representation in PyPSA-Eur

Scenario design and optimisation experiments for large-scale energy system planning using PyPSA-Eur and scientific Python tooling.

Open
04Game Development

Fool's Ascension

Roguelite card game in development, built on the Russian card game Durak. Branching 3-act runs, persistent meta-progression, and tightly-systemised encounter design in Godot.

05Creative Practice

Music Production and Sound Exploration

Explorations in musical composition and sound design, blending technical precision with artistic expression.

Open
06Side Project

OBK Gear Optimiser

A small constrained-optimisation web app for tuning gear setups in a kart racing game. A hobby project applying search algorithms to a game-balancing problem.

Open

© 2026 Ellyess Benmoufok · London