Photoacoustic Tomography (PAT) is an emerging imaging modality in which pulsed laser excitation generates ultrasound waves within biological tissue. The resulting imaging problem naturally gives rise to two coupled inverse problems: an optical problem, governing light transport and absorption, and an acoustic problem, describing wave propagation in heterogeneous media. In this project, we adopt a model-based full waveform inversion (FWI) framework that simultaneously accounts for both optical and acoustic heterogeneities. This approach enables the joint reconstruction of optical and acoustic parameters, offering the potential for improved image quality and quantitative accuracy compared to sequential or simplified methods. However, the resulting inverse problem is highly non-linear, ill-conditioned, and computationally demanding. Each inversion requires repeated large-scale forward and adjoint wave simulations, making efficient use of high-performance computing resources essential. This proposal is a continuation of our recently published work in Photoacoustics [1], where the feasibility of the proposed method was demonstrated for a two-dimensional test case. The existing implementation is fully distributed and uses matrix-free methods for the acoustic wave propagation, making it suitable for large-scale execution on distributed-memory HPC systems. The aim of this proposal is to evaluate the method at substantially larger computational scales, using large-scale two-dimensional problems and, ultimately, realistic three-dimensional models. [1] Manne Segerlund, Torbjörn Löfqvist, A multimodal adjoint-state formulation of Quantitative Photoacoustic Tomography with external acoustic sources, Photoacoustics, Volume 51, 2026, 100872, ISSN 2213-5979, https://doi.org/10.1016/j.pacs.2026.100872.