Rolling-element bearings are one of the most important components in rotating machinery, and their sudden failure can result in heavy damage to the remaining components of the bearings and nearby machines. Decreasing or minimising the effect of damage, prediction of the Remaining Useful Life (RUL) of the bearings will help to reduce the downtime cost and safety risks. The main aim of this paper is to measure multivariate data such as vibration, temperature, load, etc. and propose a data-driven RUL prediction structured methodology with the help of real-time run-to-failure vibration and Temperature data. The proposed work is divided into different parts, such as signal preprocessing, sliding-window segmentation, RUL-Labelling based on degradation and a Transfer-Learning Temporal Convolutional Network(TCN) for regression. A specific case study is conducted using the LTU-CBM run-to-failure bearing dataset. The proposed methodology demonstrates that the model achieves accurate and stable RUL predictions and performs well compared with the conventional machine learning approaches. The results show the effectiveness of temporal deep learning models for bearing prognostics applications.