About the Project
We are looking for a highly motivated PhD student to work at the intersection of machine learning, artificial intelligence, molecular modelling and thermodynamics.
Advanced molecular thermodynamic models such as SAFT-γ-Mie can provide powerful predictions of the behaviour of complex chemical systems, but their use often depends on parameters that require substantial experimental data and expert effort to determine. This project will investigate whether interpretable machine-learning methods can predict these parameters directly from molecular structure, substantially accelerating the development and application of advanced thermodynamic models.
The student will develop machine-learning models that map molecular representations and descriptors to SAFT-γ-Mie group-contribution parameters. Alongside established molecular descriptors, the project will investigate new molecular representations inspired by topological data analysis and modern machine-learning methods. Particular emphasis will be placed on models that are not simply accurate black boxes, but are interpretable, physically meaningful and able to generalise to new chemical groups.
The project will provide training and research experience in areas including:
- machine learning and interpretable AI;
- scientific machine learning;
- molecular representations and descriptor engineering;
- topological data analysis;
- physically constrained modelling;
- uncertainty quantification and model validation;
- thermodynamic modelling using SAFT-γ-Mie;
- Python-based scientific computing, data analysis and software development.
The student will be part of a strongly interdisciplinary research environment at Imperial College London, combining expertise in AI and machine learning, optimisation, molecular thermodynamics and statistical mechanics.
An important component of the PhD is its close collaboration with Eli Lilly and Company. The student will work with researchers at Imperial and industrial scientists, particularly Dr Salvador Garcia at Eli Lilly, providing the opportunity to understand how advanced AI, machine-learning and molecular-modelling methods can be translated into real industrial research problems.
The research will initially involve constructing datasets linking molecular structures, existing SAFT-γ-Mie parameters and thermodynamic information. The student will then develop and compare different molecular representations and interpretable machine-learning approaches. Predicted parameters will subsequently be incorporated into the SAFT-γ-Mie equation of state and assessed through predictions of thermodynamic properties such as phase equilibria and solubilities. The ultimate goal is to develop an end-to-end framework capable of proposing reliable parameters for previously unparameterised chemical groups.
We particularly welcome applicants from chemical engineering, chemistry, physics, mathematics, computer science, data science or related quantitative disciplines. Previous experience with Python, machine learning or computational modelling would be advantageous, but is not essential for candidates with strong quantitative skills and an enthusiasm for developing expertise in AI and machine learning.
Applicants should submit the electronic application form, submitting a CV, transcripts, a cover letter and the information of two referees through College application portal:
https://www.imperial.ac.uk/study/pg/apply/how-to-apply/apply-for-a-research-programme-/
The prospectus, entry requirements and application form (under ‘how to apply’) are available at: http://www.imperial.ac.uk/pgprospectus
Funding Notes
This is a fully funded four-year PhD studentship supported through the EPSRC Industrial Doctoral Landscape Award (IDLA), in collaboration with Eli Lilly and Company. The studentship covers tuition fees at the Home rate and provides a tax-free stipend for four years. The project will involve close collaboration with Eli Lilly, including interaction with Dr Salvador Garcia and other scientists at Eli Lilly, giving the student exposure to both academic and industrial research environments.
Voici une liste de prévisions...
Vous devez vous connecter pour voir la description complète de l'emploi et postuler.
Gérer / Mettre à niveau cet emploi vers une liste complète d'emploi.
Trouvez votre meilleure opportunité
leur dire AcademicJobs.com Envoyez vous !
