EdFitter: a hybrid course recommendation system using large language models

Educational institutions face increasing pressure to adapt their curricula to the rapidly evolving demands of the job market, and recommender systems can be important tools to guide students in selecting courses that align with their career goals. We present EdFitter, a hybrid course recommendation system that combines pretrained text embeddings for retrieval with large language models (LLMs) used offline for semantic skill extraction. The system is designed for a cold-start setting in which free-text job queries must be matched to a university’s institutional course catalog. Instead of relying on direct generative recommendations, EdFitter ranks stored course descriptions by cosine similarity and augments the output with extracted skills and matching vacancy examples. We evaluate the system in a pilot institutional deployment at Higher School of Economics (HSE) using a multi-level study with student respondents, controlled assessors, and interviews with program directors. The results indicate that users perceive the recommendations as sufficiently relevant and potentially useful to justify further development. Overall, the study suggests that EdFitter is a valuable tool for supporting course exploration and curriculum-to-labor-market alignment; we also highlight important limitations and directions for future work.