Science, technology, engineering, and mathematics (STEM) education involves a complex and diverse array of interdisciplinary knowledge structures that learners may not have in a systematic interdisciplinary framework. Therefore, when confronted with different scenarios, their manifestations of creative resilience can vary. A knowledge graph that constructs entities of interdisciplinary knowledge is advantageous for bolstering the creative resilience of learners in STEM interdisciplinary education. This manuscript presents a design scheme for a STEM knowledge graph leveraging large language model (LLM) technology, with the objective of assisting learners in improving their creative resilience and facilitating the informed transformation and enhancement of STEM education.