22MnB5 hot stamping steels are pre-coated with an Al-Si layer to prevent oxidation and decarburization during the austenitization process, which includes heating and soaking within a furnace. Above ∼577 °C, the Al-Si coating transforms into Fe-Al and Fe-Al-Si intermetallic phases because of diffusion and solidification. In the present work, a deep neural network (DNN) model was developed to predict coating intermetallic growth under various heating rates and soaking conditions. The DNN model was trained using experimental data (energy dispersive x-ray spectroscopy chemical scans) collected from our coating growth characterization work. The heating rate (r), heating time (th), soaking time (ts), and the distance of each measured point away from the coating surface (d) are defined as input parameters, and the outputs are the weight percentage of Al, Fe, and Si. The model successfully predicted the chemical composition of the training dataset, which consisted of a single-stage heating process, followed by a soaking stage at a constant temperature. To validate the model, complex two-stage heating tests were conducted, and the intermetallic coating growth was quantified. The DNN model successfully predicted the two-stage heating test coating profiles, despite a minimal error that was due to the variation in as-received coating thickness. The DNN model predicts the overall experimental trends that show the heating rate has little effect on the transformed intermetallic species, but it affects the transformation rate of the intermetallic phases. Moreover, the higher heating rate test (after 750℃) resulted in the formation of the island shaped τ1 and/or τ2 phase, while the heating rate before 750℃ does not affect the morphology of the τ1 and/or τ2 phase.