Epilepsy is a neurological disorder that affects millions worldwide, which features seizures. Though the Electroencephalogram (EEG) is the standard test for manual processing, it is expensive, time-consuming, and liable to error because of humanity’s limited endurance compared to machines’ capacity to process data continually. Even though there is “class imbalance,”where rare seizure occurrences result in poor sensitivity in clinical models. This research proposes a robust “Seizure detection framework using the TUH EEG Seizure Corpus.”(v2.0.3). In this research, we propose a Hybrid that integrates Convolutional Neural Networks to extract spatial features and Long Short-Term Memory networks to capture temporal dependencies. A significant contribution of this research is a data balance approach through under-sampling for handling the lack of seizure data. The developed model performed well, achieving 91.0% Recall and 94.7% accuracy on the validation set, considerably better than the baseline of around 30% from previous works. This shows that dealing with data imbalance is a crucial step in the process of developing effective screening devices that are of a clinical grade to help neurologists.