<p>This study addresses the problem of the high computational cost associated with fine-mesh Computer-Aided Engineering (CAE) simulations for injection molding. To overcome this challenge, it proposes an Artificial Neural Network (ANN) based approach to predict fine-mesh fill time and temperature distributions using coarse-mesh data. While fill time prediction is relatively straightforward, temperature distribution is more complex due to its dependence on multiple factors such as heat transfer dynamics and material properties. Predicting temperature solely based on coordinate and temperature values has inherent limitations, as it fails to capture spatial variations effectively. To address this issue, this study introduces gradient-based feature engineering, incorporating spatial temperature gradients as additional input features. By leveraging these engineered features, the proposed model enhances the accuracy of fine-mesh temperature predictions, improving the overall reliability of AI-accelerated CAE analysis.</p>