2026 ASEE Annual Conference & Exposition

Selecting an academic program that aligns with a student’s interests, abilities, and preferred learning style is a crucial yet often challenging decision during the early stages of college. Many students struggle to identify a program that matches their personal strengths, which can lead to disengagement, program switching, or even withdrawal from higher education. To address this issue, this research develops an Artificial Intelligence–driven recommendation system designed to assist students in selecting programs that best fit their individual profiles. The system was developed for the Department of Engineering, Data, and Computer Sciences, which offers seven specialized degree programs that share foundational subjects but differ in focus and learning outcomes.

The proposed system employs a LangChain agent that orchestrates multiple tools, including a semantic retriever, an embedding-based scoring module, and external knowledge access through Wikipedia, to evaluate each student’s knowledge areas, soft skills, personal attributes, and career aspirations. Program information and student profiles are converted into dense vector representations using OpenAI’s text-embedding-3-small model, producing 1,536-dimensional embeddings that capture semantic meaning beyond exact keywords. These embeddings are stored and searched within a FAISS vector database to enable efficient semantic similarity comparison between student attributes and program descriptions.

The LangChain agent autonomously determines when to apply embedding-based similarity calculations and when to invoke external knowledge sources, combining quantitative precision with contextual reasoning. A series of experiments conducted using real and simulated student profiles demonstrate that the hybrid LangChain-agent approach outperforms deterministic vector similarity methods alone, producing recommendations that align more closely with expert academic advising decisions. The findings highlight the potential of LLM-driven agent architectures to enhance personalized educational guidance, reduce program-selection uncertainty, and improve student retention through data-informed advising support.