Knowledge base question answering (KBQA) is designed to respond to natural language inquiries by utilizing factual information, such as entities, relationships, and attributes, derived from a knowledge base (KB). The advent of large language models (LLMs) has significantly boosted the performance of KBQA, owing to their exceptional capabilities in content comprehension and generation. In this paper, we present a Knowledge Ocean enhanced Salary Analytics (KOSA) system based on knowledge graphs and LLMs tailored to employee salary data from a public university. This system encompasses an interactive conversational interface, visualization of knowledge graphs, and advanced data analysis. By employing the framework of knowledge engineering, we enable knowledge graph modeling, Cypher (the query engine of Neo4j) reasoning, and question answering functionalities. Furthermore, machine learning algorithms are integrated to facilitate advanced features, such as salary prediction and allocation.