We are interested in developing medical AI systems that are able to perform step-by-step reasoning, in order to carefully arrive at a conclusion, whereas intermediate reasoning steps are grounded in (and verified with) high-quality medical knowledge sources.