Research Scientist/Engineer, GNNs

What We're Looking For

We are looking for a Machine Learning Engineer to take ownership of training and fine-tuning machine-learning interatomic potentials (MLIPs) for magnetic and structural materials. You will work at the intersection of modern ML and first principles simulation, leveraging our DFT datasets to feed our models and pushing MLIP architectures into new physical regimes; particularly spin dependent interactions.

What You'll Do

• Pre train and fine tune MLIPs (MACE, CHGNet, Orb, or equivalent) for solid state systems, with a focus on magnetic materials.
• Design and build DFT training set workflows, including active learning loops, convergence testing, and data curation.
• Extend existing MLIP architectures to capture spin lattice interactions.
• Build and maintain automated, reproducible workflows for dataset generation and model iteration using tools such as AiiDA, FireWorks, or equivalent frameworks.
• Work directly with materials scientists to translate physical intuition about magnetism into training objectives and dataset design decisions.

Skills & Qualifications

• PhD in physics, chemistry, materials science, or a closely related field; solid state focus strongly preferred.
• Proven hands on experience training or fine tuning MLIPs, with a clear understanding of training dynamics, loss landscapes, and generalisation behaviour.
• Experience working with DFT generated training sets and experimental material science data; understanding what makes a dataset sufficient or deficient for a given system, and being able to work with our DFT team or our experimental scientists to diagnose and close gaps.
• Strong Python skills and production quality research code; experience with PyTorch or JAX; ideally also C/C++ or Rust.
• Familiarity with atomistic simulation packages (VASP, Quantum Espresso, LAMMPS, or similar).
• Evidence of significant research impact through publications in ML for atomistic modelling, computational materials science, or related technical disciplines.

Nice to Have

• Background in long range or equivariant message passing architectures for extended systems.
• Experience with spin polarised or non collinear DFT calculations.
• Contributions to open source atomistic simulation or ML packages.
• Experience with automated workflow frameworks such as AiiDA or Fireworks.

We offer competitive salary, generous equity and benefits.

EEO & Accessibility Statement

Diffractive is an equal opportunities employer. We are committed to creating an inclusive environment for all employees and welcome applications from people of all backgrounds, experiences, and identities. If you require any adjustments or accommodations at any point during the interview process, please let us know - we will be happy to help.