Despite rapid advances in genetics and informatics, the disease trajectory and underlying genetic etiology of most childhood genetic epilepsies remains largely unknown. A genetic diagnosis can be essential to guiding treatment and care of individuals with these disorders, yet identification of a genetic cause often occurs years after onset of symptoms. A central impediment is clinicians’ reliance on manual review of data including electronic medical records (EMR) and scalp EEG to guide treatment and diagnosis of these individuals. Millions of individuals’ EEG and EMR are held in large clinical repositories across the world, yet these resources have yet to be taken advantage of at scale. There is a clear need to develop robust quantitative methods to analyze this multimodal data at scale to better understand, treat, and diagnose individuals with genetic epilepsy.In this dissertation, I develop automated, computational tools to leverage genetic, EMR, and clinical EEG to better and more efficiently diagnose these individuals and reveal the longitudinal trajectory of childhood genetic epilepsies. Specifically, I employ computational phenotyping and semantic similarity algorithms to harmonize heterogeneous clinical data to quantify phenotypic similarity and signatures of genetic epilepsies across development. I then develop a semi-automated natural language processing (NLP) pipeline to extract these phenotypes from free-text EMR, identify clinical signatures of specific genetic disorders prior to clinical diagnosis, and predict a future genetic diagnosis. Finally, I develop a pipeline to extract and clean scalp EEG to predict specific genetic diagnoses and identity novel biomarkers of disease and outcome. Overall, the methods and results of this dissertation support the use of computational phenotyping using EMR and EEG to characterize and predict genetic epilepsies across development.