Research Assistant – AI Optimisation for Materials Design

As a flagship research center in nanoscience and nanotechnology, our mission is to open and explore new frontiers of knowledge at the nanoscale, and bring value to society in the form of new understanding, capabilities and innovation, while inspiring and providing broad training to the next generations of researchers. Our values are Commitment, Collaboration and Transformation.

Our research lines focus on the newly-discovered physical and chemical properties that arise from the behaviour of matter at the nanoscale. ICN2 has been awarded with the Severo Ochoa Center of Excellence distinction for three consecutive periods and and ). ICN2 comprises 19 Research Groups, 7 Technical Development and Support Units and Facilities, and 2 Research Platforms, covering different areas of nanoscience and nanotechnology.

Job Title : Research Assistant - AI Optimisation for Materials Design

Research area or group : Theoretical and Computational Nanoscience Group

Description of Group / Project : We are building an optimisation-driven framework that (i) makes AI agents reliably operate advanced scientific software (e.g., DFT, Wannierisation, and quantum-transport codes) and (ii) uses heuristic optimisation to explore and improve materials candidates-especially 2D / vdW structures-for target electronic / spintronic properties.

The postdoctoral researcher will lead the optimisation component : formulate the design space (categorical + continuous variables), craft domain-aware operators, and couple simulation results back into the search loop. The broader goal is to deliver a reproducible, scalable optimiser that accelerates both workflow efficiency and materials discovery.

Main Tasks and responsibilities : AI4LSQUANT aims to accelerate quantum modelling by learning fast, accurate surrogates of electronic Hamiltonians. The postdoctoral researcher will develop graph neural networks based on the MACE architecture to predict Hamiltonian elements for 2D materials and van der Waals heterostructures, with a focus on spin-orbit coupling and symmetry consistency. The models will interface with existing DFT and quantum-transport workflows to validate band structures and derived properties.

Detailed tasks and Responsibilities :

Model design and training : Adapt and extend MACE-style equivariant GNNs for Hamiltonian prediction (including spin and SOC), enforcing physical constraints such as Hermiticity, sparsity, and symmetry.

Data pipeline : Curate datasets from DFT calculations (and, where relevant, Wannier / TB extractions); implement preprocessing, splits, and rigorous validation.

Metrics and validation : Define and track metrics (MAE on Hamiltonian elements, band-structure error, DOS overlap, transport-relevant figures); compare against classical parameterisations.

Uncertainty and active learning : Integrate uncertainty estimation (ensembles or evidential methods) to prioritise new calculations and improve data efficiency.

Integration : Export learned Hamiltonians to downstream simulation tools for band / transport checks; maintain reproducible pipelines (configs, seeds, provenance).

Collaboration and reporting : Work with the PI and team, contribute clean, well-tested Python code, document results, and prepare internal reports and manuscripts.

Requirements :

PhD in Physics, Materials Science, Computational Science / Engineering, Computer Science, or related.

Solid knowledge of machine learning, including graph neural networks and transformers.

Practical experience with density functional theory (setups, convergence, interpreting outputs).

Strong Python and deep-learning stack (preferably PyTorch); good software practices (git, testing).

Comfortable with Linux / HPC environments.

Nice-to-have :

Experience with equivariant GNNs (e.g., E(3)-equivariance), MACE or related message-passing models.

Familiarity with Wannierisation / TB workflows and exporting models to transport codes.

Background in spintronics or SOC-driven phenomena.

Exposure to uncertainty quantification and active learning in scientific ML.

Summary of conditions :

Full time work (37,5h / week)

Salary will depend on qualifications and demonstrated experience.

Support to the relocation issues.

Life Insurance.

Work-Life Balance and Flexibility with flexible work schedules

28 holidays per year

Flexible compensation plan : tax advantages contracting some products (health insurance, childcare, training, among others.)

International environment

How to apply :

All applications must be made via the ICN2 website and include the following :

2 Reference letters or referee contacts.

Deadline for applications : 01 / 10 / 2025

Equal opportunities :

ICN2 is an equal opportunity employer committed to diversity and inclusion of people with disabilities.

ICN2 is following the procedure for contract of people with disabilities according with article 59 of the Royal Decree 1 / 2015, of 30 of October.

#J-18808-Ljbffr