With the exponential growth of academic publications, researchers face significant challenges in efficiently discovering relevant resources for their research and development tasks. In response, recommender systems have emerged as a valuable tool, demonstrating success in various domains. In this work, we present a comprehensive architecture for an academic recommendation system based on knowledge graphs. The proposed architecture encompasses the entire process, starting from data collection, named entity recognition, knowledge graph construction to the generation of the recommendations. Our knowledge graph construction approach employs fine-tuned SciBERT, trained on SciREX, to extract key information, such as, methods and datasets referenced in academic publications. Additionally, we integrated CSO Ontology to extract topics related to the processed publications, to enhance the recommendation process. Furthermore, we propose a novel method that utilizes embeddings and similarity measures to generate personalized recommendations. To evaluate the developed prototype, we collaborated with computer science experts according to a proposed rating scale. The preliminary evaluation results confirm the system’s ability to effectively recommend relevant publications.