The recently introduced statistical learning procedure known as deep learning has proven exceptionally effective in application areas as diverse as enhancing the speech recognition on cellphones to diagnosing melanoma from photographs of skin lesions. Primarily applied to domains with exceptionally large datasets and driven by efforts from companies such as Google and Facebook to identify objects or friends in photographs, its application to other domains such as healthcare has been hindered by the lack of large curated datastores with labels appropriate for training. Or has it? Here we report our experience applying the concept of transfer learning to answer statistical questions in quantitative medicine. Specifically we review the fundamentals of transfer learning, describe how it might be applied to the "small" datasets typically provided to biostatisticians, and describe our specific experiences employing the concept on several real world clinical problems being approached at Mayo Clinic.