This document describes a neural question answering system applied to the BioASQ 5B dataset. The system uses an extractive question answering model called FastQA that is pre-trained on the large SQuAD dataset and then fine-tuned on the smaller BioASQ data. The FastQA architecture is modified to allow multiple answer spans for list questions. Experimental results show the system performs competitively on factoid questions in BioASQ and incorporates domain-specific biomedical embeddings to require less feature engineering than traditional question answering systems.