Modern manufacturing systems produce sensory data every day and methods for efficiently uncovering useful information (quickly, accurately, and with little or no human effort) within this raw data are crucial. A very promising solution in this respect is end-to-end data processing. It allows capturing complex tasks through a single complete model. Recent progress within the deep-learning community has shed light on the implementation of end-to-end data-processing techniques in manufacturing. These techniques take advantage of the learning ability of Artificial Neural Networks (ANNs) to extract essential system features from the inputs. Based on this information, a simple model is constructed to complete modeling. This thesis focuses on neural-computing-based approaches to processing industrial sensory data. Predictive machine maintenance, work-in-progress on product-quality prediction, and fast data modeling are selected to show how neural computing can be used to uncover useful information from industrial sensory data. In this thesis, neural computing refers to ANNs-based methods for solving specific problem-modeling tasks.