The increasing complexity and volume of extracellular neural data generated by modern silicon-based probes present significant storage, transmission, and analysis challenges. Spike sorting algorithms have been developed to cope with this data deluge, but innovative approaches are required to compress the data while maintaining information fidelity efficiently.This project explores the application of deep learning techniques for compressing extracellular neural data, leveraging the hierarchical representations learned by convolutional neural networks to achieve efficient compression without compromising relevant information.We trained a convolutional neural network architecture on extracellular neural data to learn compressed representations of the signals with training data consisting of samples of 128x128 dimension extracted from extracellular recordings. The models were optimized to minimize reconstruction error while reducing the dimensionality of the data. To achieve this, unsupervised and supervised learning approaches were used.The model training consisted of several steps, in the different steps, different training methods were applied to the base model to increase its final performance and efficiency.By further fine-tuning the optimizing policy of the model, a highly efficient, promising compression method can be developed for spike sorting in the future, enabling on-device compression of electrophysiological data without major compromises in the sorting process.