PDRA in ML and NL Processing for Materials Science

The Role

Applications are invited for a Research Associate in Machine Learning and Natural Language Processing for Materials Science.

The project focuses on building a database and on combining advanced ML and NLP approaches to accelerate discovery in Molecular Solid Solutions (MoSS). Data will be gather collectively during the course of the project. NLP techniques will be used to automatically extract key data points from published papers and reports, building and enriching a comprehensive MoSS database. Machine Learning will then be applied to this database, alongside quantum chemistry methods, to predict MoSS formation, model dopant effects on crystal structures, and identify optimal dopants for tuning host compounds. The project is highly multidisciplinary, linking computational and physical chemistry, physics, and computer science at Durham University and partner institutions.

The successful applicant will be expected to design and implement NLP pipelines for literature mining, and to select, adapt, and develop ML models, predictive models in PyTorch, and interpretable methods such as SHAP and causal inference. This dual approach will both expand the data resource and provide powerful predictive insights into structure–property relationships. The PDRA will also produce python code for advanced scientific data processing. This will then be implemented in the data pipelines of the project. The applicant will also serve as data champion to the project.

The postholder will be expected to contribute actively to high-quality research outputs, including publications, software tools, and follow-on grant proposals.

Advert information