Speaker: Emma Lejeune

Title: Reproducibility first computational mechanics

Abstract: Large datasets combined with modern computational methods offer unprecedented opportunities to model, analyze, and understand complex systems. In computational mechanics and mechanobiology, the past decade has seen rapid growth in data science and machine learning applications. Yet a fundamental question remains: how do we know when a computational method is effective, generalizable, and reproducible? In this talk, I will highlight recent work organized around two complementary components. First, I will describe our efforts to construct benchmark datasets for the solid mechanics community, beginning with Mechanical MNIST and extending to large-scale simulations of deformation and fracture, and introduce recent validation studies assessing the ability of large language models to generate and verify finite element analysis code. Second, I will discuss our work developing open-source biomedical image analysis tools that make the research pipeline more reproducible and extract richer information from experimental data. Throughout, I will emphasize open-access datasets, open-source software, and community engagement as the foundation for reliable and reproducible computational science.

About the Speaker: Dr. Emma Lejeune is an Assistant Professor of Mechanical Engineering at Boston University. She earned her Ph.D. in Computational Mechanics from Stanford University in 2018 and has been at BU since 2020. Her research sits at the intersection of computational mechanics, scientific machine learning, biomechanics, and mechanobiology, developing benchmark datasets, open-source software, and validation frameworks that help the community evaluate whether methods are truly reliable and reproducible.