Gynecological conditions affect women’s health on a world-wide scale which requires timely and effective diagnosis. This research study investigates the fine-tuning of Bidirectional and Auto-Regressive Transformers (BART) for the purpose of automated diagnosis based on patient symptoms. We will attempt to develop a BART model based on a curated dataset of gynecological conditions which link out relevant symptoms to diagnoses by providing an accurate mapping of these diagnoses and their relevant symptomatic characteristics and treatment indications. We will use transfer learning and supervised fine-tuning protocols to evaluate the performance of this model finally integrating our proof of concept into a Streamlit based web application for public use. Our experiments confirmed the model was effective at providing accurate diagnostic predictions, along with substantial gains in accuracy and recall including f1score. The results indicate that models based on transformers such as BART can be considered useful in the clinical decision making process, which may also reduce delays in patient diagnoses and improve standards of care.