As contribution to the 2023 George B. Moody challenge, we-team 'BrAInstorm'-aimed for fusing semantic features based on medical knowledge with an end-to-end residual neural network to predict the secondary neurological outcome after successful resuscitation. More precisely, we fused numerical (e.g. age) and categorical (e.g. gender) information as well as features extracted from biosignals: We extracted absolute and relative power bands, coupling, and coherence from standard electroencephalography (EEG) frequency bands. To investigate the interplay between heart and brain, we computed deceleration capacity (DC) from electrocardiograms (ECGs). In contrast to these semantic features, we adapted a residual neural network based on agnostic features which are derived from the training data. The network architecture was originally developed for classification of ECGs and was adjusted to the challenge EEG data. The best metric scores were reached using only the neural network, demonstrating the complexity of outcome prediction and effectiveness of end-to-end methods. We received a challenge score of 0.57 ± 0.15 during 5-fold cross validation on training data and 0.448 on the hidden validation data. On the hidden test data we received a final score of 0.68 (rank 8 of 36).