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 cuttingedge AI-powered tutoring for STEM education.