We advance AI and integrate it with engineering knowledge to accelerate the discovery, prediction, design, and optimisation of sustainable chemical and biochemical systems, from molecular innovation to smart manufacturing.
We combine physical models with machine learning to predict chemical and biological process behaviour. Our work investigates how hybrid models can be developed with limited data, adapted between systems and scales, and used as a foundation for process digital twins.
We develop methods that discover mathematical relationships and candidate physical explanations from data. By combining symbolic regression, engineering knowledge and experimental design, we investigate how interpretable models can support scientific understanding and process development.
We turn industrial measurements and operating histories into models of product quality, batch behaviour and equipment condition. Our work includes soft sensors, interpretable quality models, reconstruction of batch operations and predictive maintenance using limited labelled data.
We connect predictive models with decisions about process operation. Our research investigates flexible operating regions, optimisation under uncertainty and intelligent control strategies for batch processes and biological production systems.
We connect biological behaviour with reactor conditions and production performance. Our work brings together cell metabolism, fermentation, mixing and light-dependent growth to investigate bioprocess design, prediction and scale-up.
We develop AI methods to generate molecules with targeted properties and assess their practical accessibility. Our work spans surfactant design and cost-aware molecular screening, connecting candidate generation with the challenges of making or purchasing promising compounds.
Complex systems rarely offer complete mechanisms or perfect datasets. Our research develops methods that work with these limitations while retaining a connection to the underlying engineering.
Which mechanisms explain the observations, and how can interpretable models help reveal them?
How can models remain useful when data are scarce, conditions change or a process moves to a different scale?
Which experiments, operating conditions or molecular candidates offer the most useful next step?
The balance between physical models, machine learning and optimisation depends on the system, the available evidence and the intended use. Four principles run through our work.
Material balances, kinetic relationships and process constraints provide structure. Machine learning represents behaviour that is difficult to describe mechanistically.
Experimental design, transfer learning and analysis of operating histories help extract useful information from limited or imperfect observations.
We investigate model assumptions, interpretation and uncertainty using experimental datasets, industrial records and simulation benchmarks.
Models become tools for process design, quality prediction, scale-up, scheduling and control, with methods chosen around the decision being made.
Common modelling and decision-making challenges connect applications that can otherwise look very different.
Reaction kinetics, mechanism discovery and catalytic process modelling.
Fermentation, cell cultures, microalgae and bioreactor scale-up.
Batch manufacturing, quality prediction and surfactant design.
Biohydrogen, anaerobic digestion and optimisation of energy-related processes.
We welcome conversations about research collaborations, industrial challenges and opportunities to work with the group.