Alberto Ceselli

Associate Professor at the Department of Computer Science of University of Milan, Italy

Alberto Ceselli is a Full Professor in the Department of Computer Science at the University of Milan. He received his Ph.D. in 2006 and spent research periods as a visiting scholar at leading European institutions, including TU Berlin, Politecnico di Milano, INRIA Sophia Antipolis, RWTH Aachen, Université Paris XIII, and ESSEC Business School. His research focuses on computational integer programming, prescriptive data analytics, and bridging optimization with data-driven decision-making. He is actively engaged in technology transfer, collaborating with industry on applications in network optimization, smart cities, and digital industry. He also coordinates the department’s relations with its industrial advisory board.

Talk Title: From Models to Decisions: Designing and Validating Optimization-Based Decision Systems

Abstract: From Models to Decisions: Designing and Validating Optimization-Based Decision Systems Operations Research has developed, over decades, powerful methodological foundations for modeling and solving complex decision problems. Yet strong model and solver performance does not automatically translate into operational use or impact. In fact, the operational performance of an optimization method depends not only on the formulation and algorithms, but also on the data, the representation of uncertainty, the surrounding workflow, and the way recommendations are interpreted and implemented. I argue that increasing the impact of optimization requires moving from isolated models and algorithms toward the integrated design of the full chain from data to decision. Drawing on both academic and industrial case studies, I will examine how interactions with data pipelines, software components, and human decision-makers affect model formulation, algorithm design, validation, and implementation, including instances in which technically strong solutions were overridden, substantially revised, or failed to produce the expected operational effect. The discussion will distinguish model-level performance from the acceptance, implementation, and operational effects of optimization recommendations. It will also identify research opportunities concerning data integration, model validation, interactive optimization, and the incorporation of user feedback. The objective is to show how the methodological strengths of Operations Research can be retained while optimization is embedded in decision processes that can be evaluated in practice.

Website: Personal webpage