Lattice Boltzmann methods (LBM) are an established mesoscopic approach to the simulation of a diverse range of transport problems, most prominently flows described by Navier-Stokes (NSE) and reaction-advection-diffusion equations. Their algorithmic structure renders LBMs uniquely suited to massively parallel high performance computation (HPC). LBMs have successfully been used in the context of computational fluid-structure interaction (FSI), primarily by coupling external structural solvers via immersed boundary methods (IBM). The homogenized lattice Boltzmann method (HLBM) [1] is a recent approach to modeling flows with dynamic boundaries that considers fluid and solid regions as a unified, time-dependent porous medium. Compared to established approaches, the HLBM offers advantages for HPC due to its local, non-interpolated nature, straightforward coupling to arbitrary structural models as well as suitability for FSI with massive self contacts. This presentation discusses a novel approach to the simulation of arbitrary FSI problems with two-way coupled deformable boundaries using HLBM in combination with various structural models ranging from simple ODEs up to fully discretized solid continuum mechanics. Numerical validations w.r.t. academic reference cases are presented, including a model of a mechanical heart valves [2] and filament oscillations. The proposed performance-focused approach is of interest for large-scale ensemble simulations aimed at uncertainty quantification of flows in biomedical and common industrial applications. Showcased applications range from unsteady hemodynamics in moving vessels over a fully coupled model of the human vocal fold to centrifugal pumps and continuous stirred tank reactors (cf. the partner abstract on wall modeled LES using HLBM by Bukreev et al.). Finally, the implementation and parallel performance on heterogeneous Multi-GPU clusters using the open source LBM framework OpenLB [3] will be discussed. [1] Simonis et al. ”Homogenized lattice Boltzmann methods for fluid flow through porous media – part I: kinetic model derivation”. Under review. (2024). DOI: 10.48550/ARXIV.2310.14746 [2] Stijnen et al. ”Evaluation of a fictitious domain method for predicting dynamic response of mechanical heart valves”. In: Journal of Fluids and Structures. (2004). [3] Krause et al. ”OpenLB—Open source lattice Boltzmann code”. In: Computers & Mathematics with Applications. (2020). DOI: 10.1016/j.camwa.2020.04.033