International Journal of Scientific Research and Engineering Development

The rapid growth of online career and learning resources has made it difficult for job seekers and professionals to identify the skills, roles, and learning paths that best match their goals. This paper presents SkillGraph, a multi-agent architecture for AIpowered career recommendation that combines a skill-career knowledge graph (KG), graph neural network (GNN) based ranking, and large language model (LLM) driven agents coordinated by a central orchestrator. Unlike single-model recommendation pipelines, SkillGraph decomposes the recommendation task across specialized agents responsible for profile and skill extraction, knowledge-graph retrieval, GNN-based ranking, career-path planning, and feedback incorporation, communicating through a structured message protocol and a shared short-term, long-term, and vector memory layer. We describe the system architecture, agent responsibilities, coordination algorithms, and deployment design, and report an experimental evaluation against content-based, collaborative-filtering, single-agent LLM, and GNN-only baselines. In our evaluation setting SkillGraph achieves higher Precision@10, Recall@10, and NDCG@10 than all baselines, and an ablation study shows that the orchestrator and shared memory components contribute the largest gains to task completion rate. We discuss the strengths, limitations, and privacy considerations of the proposed design and outline directions for self-learning and dynamically composed agent teams.

Jahnavi Somaraju, V. Guru Thrinath, S. R. Bhavishya, S. Bhavya, Y. Jahnavi, "SkillGraph: A Multi-Agent Architecture for AI-Powered Career Recommendation Using Knowledge Graphs and Graph Neural Networks" International Journal of Scientific Research and Engineering Development, V9(4): Page(485-498) May-June 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.