This collaborative effort seeks to integrate seismological, geodetic, experimental, and geological constraints for the Japan, New Zealand and Japan natural subduction zone laboratories into numerical models to advance our understanding of megathrust earthquakes. Forward models and a new numerical modeling framework for data assimilation will be deployed to get closer to versatile tools for data-driven, physics-based hazard assessment. The focus is on the evolution of fault stress and strength over a range of spatio-temporal scales, quantifying uncertainties and sensitivity to parameters. This will allow formulating best strategies for inferring relevant parameters from data in the presence of ambiguous physics. All code will be made publicly available along with cookbooks and tutorials, and a networked effort will establish new, quantitative links and leverage individual efforts greatly. FRES funding will support a growing community of solid Earth geodynamicists who want to deploy their models in a hazard and monitoring context. A focus will be on training and sharing material for interdisciplinary computational geoscience efforts, from undergraduate to post-doc and practitioner level. Project participants will develop sustainable pathways for participation and work to enhance representation and inclusion in the geosciences by providing new pathways of entry based on modeling and remote sensing to complement field-based approaches.