Handwritten Text Recognition (HTR) has continued to be a daunting task particularly when it comes to the Indic scripts like Gujarati that are very complex in nature in the form of ligature, modifiers and compound characters. The existing traditional optical character recognition (OCR) based systems, mostly formulated to be able to recognize Latin based scripts, do not incorporate the structural complexity of Gujarati hand writing, which is why these OCRs have low recognition rates. The more recent developments in deep learning proposed hybrid networks that combine convolutional with recurrent neural networks to resolve these issues. We introduce a Convolutional Recurrent Neural Network (CRNN) model in this work, which involves Convolutional Neural Networks (CNN) to extract hierarchical features, Bidirectional Long Short-Term Memory (BiLSTM) to acquire context in a sequence, and Connectionist Temporal Classification (CTC) loss to be able to do transcription without alignments. The style of the architecture also removes the necessity of explicit character segmentation, which is notoriously challenging in Gujarati because of overlapping as well as cursive strokes. In order to test the suggested model, we use a handwritten Gujarati words and image of lines curated dataset, augmented with data-related methods in order to enhance generalization. The performance of the model is compared to the traditional CNN and LSTM baselines with Character Error Rate (CER) and Word Error Rate (WER) performance measures. Experimental findings prove that CRNN is much more effective than the current approaches, which results in strong recognition even with damaged handwriting and complicated ligatures. The suggested solution, therefore, offers a scalable and stable framework in terms of developing Gujarati handwritten text recognition and its potential uses include digitization, educational applications, and archiving.