Abstract:

Educational AI systems tend to have a hard time to combine exact mathematical reasoning with adaptive, context-dependent explanations. This paper introduces STEM Bot, a multimodal educational assistant that combines symbolic computation (through SymPy), large language models (LLMs), and FAISS-based semantic retrieval into a single Streamlit interface to accommodate text, image, and PDF inputs. Through application of hybrid neuro-symbolic reasoning, STEM Bot improves the mathematical computation precision and the science/social studies explanation contextual relevance. Validation on university-level STEM datasets and user studies shows improvements in precision, explainability, and learner engagement over isolated LLM approaches. These outcomes place STEM Bot as a strong candidate for cutting edge AI-powered tutoring for STEM education