This paper describes an effort to apply current structural, thermal, and fluid reduced order modeling methodologies (ROMs) to multi-disciplinary interaction problems that are of interest to the aerospace industry. The study focuses on a stiffened panel configuration, originally designed by the Boeing Company, that is subjected to combined pressure and heat flux loads from a Mach 7 flow. The nonlinear structural / thermal ROMs rely on the identification of suitable bases to accurately represent the panel to anticipated loading. The structural bases are a combination of linear free vibration modes, dual modes that capture the in-plane response, and enrichment modes that capture the structural response due to thermal loads. The thermal bases include temperature distributions that are derived from the dominant bases in the structural ROM and enrichments that capture the effect of the anticipated fluid loading. The fluid surrogates are generated as the interpolation of a set of fitting points that sample a parameter space of interest. The ROMs and surrogates are generated with no information about the full order coupled response. The ROMs are verified using the corresponding full order models, both in an individual as well as in a coupled setting. The coupled panel deformation, which is predominantly driven by the thermal loading, is found to exceed assumed training bounds for the fluid surrogate in a number of cases. However geometric stiffening of the panel results in a decreased sensitivity to pressure loading, thereby producing reasonable agreement in the coupled response when using the full and reduced order fluid models. Significant differences, stemming from the type of elements used, were noted in the two full order thermal models employed in the studies. This produced drastically different coupled responses when using the structural / thermal ROMs and the full order simulations. The results highlight the potential of reduced order approaches in accurately modeling the multi-disciplinary coupled response of complex built-up structures and the significant computational gains offered over the full order models, particularly in the context of repeated simulations or unsteady response studies.