PhD on automated feature discovery and data-driven modelling for chemical processes

Within EL4CHEM - Efficient Learning for Chemical Applications, the researcher will develop automated methods to identify the most informative features and construct reliable data-driven models for complex chemical and pharmaceutical processes. Industrial process datasets typically contain many potential inputs, ranging from operating conditions and sensor measurements to material properties, molecular descriptors and engineered process variables. At the same time, experimental data are often scarce, noisy and expensive to obtain. Selecting the right information therefore becomes as important as selecting the modelling method itself. The PhD will investigate automated feature generation, feature selection and model identification methods that can determine which variables and representations are most relevant for a given prediction or modelling task. The work will combine modern machine-learning techniques with chemical-engineering knowledge to develop models that are accurate, interpretable and robust under limited-data conditions. Research topics may include:

• automated feature generation and selection for process and product data;
• sparse and interpretable machine-learning models;
• nonlinear feature interactions and dimensionality reduction;
• automated comparison and selection of data-driven model structures;
• incorporation of physical and chemical knowledge into feature-selection workflows;
• uncertainty and robustness of selected features and models;
• explainable AI methods to identify the physicochemical and process variables governing model predictions;
• development of automated modelling workflows that can be applied across different EL4CHEM industrial use cases., You will work in an international and multidisciplinary environment at the intersection of chemical engineering and artificial intelligence, with opportunities for scientific publication, collaboration with industrial partners and development of new data-driven modelling methodologies with direct industrial relevance. For more information please contact Prof. dr. Mumin Enis Leblebici, mail: . Where to apply Website, You will work in an international and multidisciplinary environment at the intersection of chemical engineering and artificial intelligence, with opportunities for scientific publication, collaboration with industrial partners and development of new data-driven modelling methodologies with direct industrial relevance.

We are looking for a candidate with a background in chemical engineering, process engineering, applied mathematics, data science, computer science or a related field. Experience with Python, machine learning, statistical modelling, feature selection, optimisation or process modelling is an advantage. A strong interest in combining machine learning with chemical-engineering problems is essential. The candidate should be motivated, independent and interested in interdisciplinary research involving both methodological development and industrial applications. We offer a full-time research position for one year, with the possibility of extension up to four years, depending on performance and available funding., Research Field Engineering Education Level Master Degree or equivalent Languages ENGLISH Level Good Languages ENGLISH Level Mother Tongue Research Field Engineering » Chemical engineering Years of Research Experience None, We are looking for a candidate with a background in chemical engineering, process engineering, applied mathematics, data science, computer science or a related field. Experience with Python, machine learning, statistical modelling, feature selection, optimisation or process modelling is an advantage. A strong interest in combining machine learning with chemical-engineering problems is essential. The candidate should be motivated, independent and interested in interdisciplinary research involving both methodological development and industrial applications.

We offer a full-time research position for one year, with the possibility of extension up to four years, depending on performance and available funding.