Cardiac arrest is a medical emergency that can have devastating consequences, including brain damage and death. Early and accurate prognosis of cardiac arrest patients is crucial for determining the appropriate treatment and management strategies. One promising approach to predicting the prognosis of cardiac arrest patients is through the analysis of electroencephalogram (EEG) readings. Electroencephalographic monitoring is cheap and easy to perform, however it fails in a few aspects, it is susceptible to sedative drugs, metabolic derangements, lack of universally recognized classification system, and hypothermia. We aim to use a deep neural network with Bidirectional LSTM architecture to predict patient outcomes. The model is binary classifier that returns whether the outcome was Good (CPC =1,2) or Poor (CPC=3,4,5). This is because we need to model both spatial and temporal features. The BLSTM component aids in capturing temporal dynamics in both forward and backward directions. Using a 70-30 split for training and testing, our model received an overall accuracy of 0.833 and a F-measure of 0.795, with a specificity of 0.904 and a sensitivity of 0.666. Further tweaking of the hyperparameters is needed for a higher performance.