Epilepsy, affecting over one percent of the global population, is a debilitating neurological disorder marked by seizures, which are coordinated firings of neuronal populations in the brain. These seizures manifest as symptoms ranging from uncontrolled automatisms to full-body rhythmic jerking episodes. Despite advances in epilepsy care, 40% of patients continue to have seizures after failing at least two medications and invasive surgeries. The qualitative analysis of intracranial EEG (iEEG) data, dependent on the brain state, remains a significant challenge. Biomarkers of epilepsy vary during baseline (interictal) and seizure (ictal) epochs, as well as across wake and sleep stages. Signal processing and machine learning methods offer the potential to automate the analysis of this data, enhancing the interpretation of iEEG for localizing epileptic foci and monitoring seizure severity over time. In this dissertation, I collate data across large, multi-center cohorts of epilepsy patients to address three specific aims. First, I investigate whether electrographic abnormalities during sleep stages vary from those during wake stages. I also characterize how choice of pre-processing for iEEG networks can affect downstream analyses. Second, I assess if a seizure severity score, composed of clinical and EEG features, can predict the response to therapy. Third, I explore whether properties of seizure onset and spread can be preserved in a deep learning model. Overall, this dissertation supports the implementation of quantitative iEEG analysis in epilepsy evaluation and treatment, offering new insights and tools for improving patient outcomes.