This site is optmised for modern browsers - e.g. Google Chrome, Mozilla FireFox or Microsoft Edge.


Correcting low-fidelity solvers with Graph Neural Networks

Supervisors: Jonathan Rayner, Nicolas Durrande
Location: Cambridge, UK. Possibility to work from home 2 days/week.
Duration: Up to 3 months. Fixed term contract with flexible starting date.
Reward: Monthly salary of £2000 for undergraduate students, £3000 for graduate students. Holiday allowance in line with the rest of the team.

Project description

Making accurate numerical simulations of complex physical systems such as electric motors typically requires detailed multi-physics finite elements simulations that are computationally intensive. In many cases, it is however possible to use lower fidelity approaches that can approximate the predictions of the complex solver for a fraction of the computational cost.
The aim of this project is to train Graph Neural Networks (GNNs) to correct for the errors observed in low fidelity solvers. More precisely, we want to investigate the ability of GNNs to approximate high fidelity solutions when given the low fidelity field obtained with a faster method. The summer placements will consist of implementing various GNNs and in studying the resulting speed-up vs accuracy trade off compared to a slow but higher fidelity solver.

Qualification and preferred skills

This position is primarily targeted at PhD students in Machine Learning. Applicants with other levels of qualifications will nonetheless be considered. We don’t expect or require any prior experience with electric motors.

The skillsets that we see as relevant for the placement are:
• Experience with deep learning frameworks. We tend to use PyTorch and we've had a good experience with PyTorch Geometric on previous GNN projects.
• Knowledge of the GNN literature, or ability to quickly get up to speed. We will do some data preparation for you, so it's great if you're able to contribute right away to the core methodology.
• Good communication skills. We'd love to regularly hear back from your findings so that the team can learn from you.
• Good scientific practices around benchmarking and reproducible research.

About Monumo

Inspired by the environmental challenge and driven by the desire to solve problems long considered too difficult, Monumo is a pioneering company at the forefront of the electrification journey. Located in central Cambridge and Coventry, we are a dynamic team of 30 professionals including motor engineers, physicists, computer scientists, machine learning experts, and entrepreneurs. Our mission is to revolutionise electric vehicle motors by developing innovative motor design technologies that leverage scalable numerical simulation tools and artificial intelligence. At Monumo, we strive to push the boundaries of what's possible, creating next-generation powertrains by combining human ingenuity with deeptech intelligence.
We are excited to offer summer placement opportunities for passionate undergraduate and graduate students who are eager to contribute to the electrification revolution. This internship is an excellent chance for students to apply their academic knowledge in a real-world setting, working alongside our multi-disciplinary team of experts in a cutting-edge field.

Application Process

The position is open to undergraduate or graduate students in engineering, mathematics, physical sciences, computer science or a related field.

Please apply for the Monumo Summer Internship Program here, including your resume and a cover letter explaining how you believe you can contribute to Monumo's journey.
Join us at Monumo for an exciting summer of innovation and learning as we drive towards a more sustainable future!