In this study, the food freezing process was investigated using a deep learning model based on a Long Short-Term Memory (LSTM) network, with a focus on reducing the computational time required for CFD simulations. Initially, the key features that directly influence heat transfer within the freezing chamber were analyzed and extracted from CFD simulation data. Based on this dataset, a deep learning model was developed to predict the core temperature evolution and freezing time of food products during both the cooling and subfreezing phases. The model was subsequently validated by comparing its numerical predictions with the data obtained from CFD simulations of salmon fillets in a large-scale freezing chamber. The results demonstrate that the deep learning model achieves prediction accuracy comparable to that of CFD simulations, while requiring significantly less computational cost.