Abstract: This paper proposes an intelligent assistant application program for railroad engineering supervision based on knowledge graph, which innovatively combines knowledge graph and deep learning to provide intelligent support for railroad engineering supervision. First, the knowledge graph in the field of railroad engineering is constructed through the steps of knowledge extraction, knowledge fusion and knowledge storage, including named entity recognition based on Bi-LSTM-CRF and entity relationship extraction based on dependency syntactic analysis, and is stored using Neo4j graph database. Then, deep learning models (e.g., BERT) are utilized to achieve intent recognition and question similarity matching, which effectively improves the accuracy and efficiency of railroad engineering regulatory Q&A. The experimental evaluation shows that this intelligent assistant application can significantly reduce the labor cost and improve the supervision efficiency, and has high application value and promotion potential in the field of railroad engineering supervision. This study not only provides intelligent support for railroad engineering supervision, but also promotes the application of knowledge graph and deep learning technology in the field of engineering supervision.