Deep learning is widely used in image recognition, speech processing, and other tasks. Text recognition from handwriting is still a challenging task due to it being associated with the inherent variability of writing styles and languages as well as the numerous representations that handwritten characters can have. Moreover, recognizing cursive writing remains an ongoing challenge due to the inherent fluidity and interconnected nature of cursive characters. This research proposes a two-stage architecture using AlexNet and bidirectional LSTM for feature extraction and language detection, respectively. When compared to other deep learning architectures such as WiRITE, VGG16, ResNet18, the proposed two-stage model has shown superiority in improving the system’s accuracy of up to 96% on the task of handwriting recognition. This research aims to capitalize on deep learning in building a productive handwriting recognition system.