Over the last several decades, electronic medical record systems have been brought into use in hospitals and clinics across the US. These systems contain volumes of rich (if noisy) data on patients’ demographics, diagnoses, vitals, lab results, medicines prescribed, procedures performed, and (with increasing frequency) genotypes. This growing pool of data has massive potential utility in tasks such as identifying risk factors and constructing predictive models. However, this type of analysis has yet to transform day-to-day clinical practice. The goals of my research are to develop novel methods that leverage this accumulated data both to identify previously unknown population-wide relationships between items of note in the medical record (such as unidentified risk factors for a disease of interest) and to enable patient-specific risk assessment for use at time of care by medical practitioners.