The implications of sea level fluctuations, mostly attributed to global warming, pose significant challenges for naval operations, navigation, coastal economies, and infrastructure resilience. Accurate sea-level prediction is vital to mitigate these challenges, particularly in areas prone to tidal flooding and infrastructure damage. This study introduces a cutting-edge deep-learning model, iTransformer, for sea level prediction, utilizing six months of data with four months allocated for training and two months for validation and testing. The dataset used originates from Singaraja, Bali, Indonesia, and projects sea levels over a 14-day timeframe. Our findings reveal that iTransformer outperforms both Temporal Convolutional Networks (TCN) and Transformer models in terms of prediction accuracy and has a significant amount of short-term computational efficiency. iTransformer achieves the prediction score with an RMSE of 0.0041 for the 24-hour, 0.0415 for the 48-hour, and 0.0455 for the 96-hour. MAE of 0.0489 for the 24-hour, 0.0323 for the 48-hour, and 0.0368 for the 96-hour, and R2 scores of 0.9713 for the 24-hour, 0.9882 for 48-hour, and 0.9863 for the 96-hour prediction windows, respectively. Furthermore, iTransformer demonstrates lower computational times, requiring only 0.1 minutes for a 24-hour window and 0.3 minutes for a 48-hour window. These results underscore iTransformer's potential as a robust model for sea level prediction and indicate that its application could be extended to multivariate datasets to enhance performance.