Aims: This work aims to improve the accurate prognosis in post-cardiack arrest patients, ensuring adequate and timely treatment, by extracting a set of relevant features from EEG signals and apllying Machine Learning techniques. Methods: Our initial approach involves the development of an algorithm, which includes various time-domain feature extraction techniques. These features capture different aspects from EEG signals, allowing for a more robust representation of underlying brain activity. We extract a total of 200 features, including characteristics from individual EEG channels acquired from time-windows, and EEG patient-level data, all used to train a Random Forest as a starting point Machine Learning model for prognostic prediction. The rigor of our work and the novelty of our approach lie in the comprehensive set of features and their effective combination for the task of predicting post cardiac arrest prognosis. Results: Our results indicate that feature extraction from EEG signals shows promise for prognosis prediction. Feature extraction is likely to be a deter-mining factor in the final competition results. Our UFC_MDCC Team achieved an official score of 0.43 using Random Forest, with no hyper-parameter optimization and using only time-domain features combined with patient clinical data. On development environment, after a 5-fold cross-validation, we achieved the following average metrics: 78.25% Accuracy, 79.66% Precision, 88.05% Recall, 83.54% F1-score, 83.60% AUC-ROC, 0.3256 Challenge Score. Conclusion: The continuity of this work is promising and has significant chances of contributing to improve the prognostic capabilities from the EEG signals, thus allowing adequate and timely treatment for patients who suffered cardiac arrest. Our preliminary results highlights the potential for improvement by exploring other signal domains, such as frequency and time-frequency, optimizing and combining more robust models (RNN) to deal with time-series data as is the case of the EEG exam.