Predicting the physical and chemical behaviour of active pharmaceutical ingredients using advanced thermodynamic models. Combining SAFT-γ Mie and COSMO-SAC frameworks to predict solubility, phase diagrams, and partition coefficients — without expensive experimental campaigns — directly supporting drug formulation and design.
Ionic liquids offer a powerful route to improve the solubility, stability, and bioavailability of drug molecules. This research develops thermodynamic models tailored to IL-APIs, applies computer-aided design to identify optimal ionic-liquid candidates, and delivers the open-source platform openAPI-ILDesign for the community.
Bridging physics and data science with graph neural networks, neural networks, and evidential deep learning to predict molecular and salt properties. A standout contribution is a GNN that predicts COSMO-derived molecular descriptors directly from structure — dramatically accelerating property screening.
Electrolyte solutions underpin everything from batteries to biological fluids, yet their complex ion interactions make them notoriously difficult to model. This work develops next-generation equations of state — Binding DH, Binding eSAFT-VR Mie — alongside novel models for electrical conductivity and ion pairing, advancing how we understand ions in solution.
The computational backbone of the group's work: rigorous conformer sampling with ORCA, CREST, Gaussian, and RDKit, coupled with COSMO quantum-mechanical calculations. These workflows generate the molecular-level inputs that feed the thermodynamic and machine-learning models across every other theme.
Asphaltene precipitation can block wellbores and pipelines, causing costly production failures. This research applies PC-SAFT to predict asphaltene phase equilibria, models multiphase flow in pipes, and predicts deposit thickness on wellbore surfaces — all underpinned by careful PVT characterisation of reservoir fluids.