Traditional knowledge construction has the problems such as strong dependence on domain experts, long construction cycles, and high demand of professions, so in industry human-machine coupling is usually adopted to improve the efficiency of knowledge construction, but still dependent largely on domain experts. To solve this, iFLYHealth has proposed an automated knowledge building and evolving technology based on deep learning. It provides the machine with the abilities to learn domain knowledge actively, to extract candidate knowledge from massive medical texts automatically, and to integrate them into medical knowledge graph, so achieves a double cycle of knowledge extracting and knowledge updating, realizes autonomous construction of a comprehensive, authoritative, professional medical knowledge base, and forms a new efficient and accurate mode of medical knowledge construction.