Vat photopolymerization (VPP) is an additive manufacturing process in which polymer resin is cured layer by layer under UV light, enabling high-resolution applications such as chip packaging. Accurate modelling of light-material interactions is critical to achieving micron-scale precision. Current approaches largely rely on volumetric light models with one-way coupling to curing dynamics, which fail to capture complex optical phenomena, including the evolving optical properties of resin during curing. In this work, we present a GPU-accelerated Finite Element Ray Tracing (FE-RT) framework bidirectionally coupled to a chemical photocuring model. The resin is discretised using finite elements, allowing piecewise-linear representation of the gradient refractive index within the material. Our simulations reveal significant differences in 3D light intensity fields and curing profiles compared to conventional models, demonstrating the importance of capturing the coupled opto-chemical dynamics. This approach provides a high-fidelity predictive tool for next-generation VPP applications requiring micron-level accuracy.