Open Access Open Access  Restricted Access Subscription Access

AutoHire Coach: An Agent-Orchestrated Retrieval-Augmented Framework for Context-Aware Career Preparation

Shyamala Nagajyothi, S. Ashwanth, R. Saketh, G.Akshay Kumar

Abstract


Modern job preparation platforms often rely on static resume parsing or generic large language model prompting, which limits contextual grounding and long- term personalization. This paper presents AutoHire Coach, an agent-orchestrated retrieval-augmented framework designed to perform semantic skill gap analysis, generate personalized project roadmaps, and produce context-aware interview ques- tions aligned with specific job descriptions. The system integrates a FastAPI backend, a React-based frontend, and an embedded Qdrant vector database to enable cosine similarity-based retrieval over dynamically growing knowledge collec- tions. A Llama 3.3 70B model accessed via cloud API performs skill extraction and reasoning, while BAAI/bge-small-en-v1.5 embeddings (384-dimensional) sup- port efficient semantic search. Unlike single-pass prompting systems, the proposed architecture combines retrieval, web search augmentation, GitHub code parsing, and cross-session knowledge persistence to enhance contextual consistency. Ex- perimental evaluation across multiple job descriptions and candidate profiles indi- cates improved relevance and stability of generated outputs when compared to a non-retrieval baseline, while maintaining acceptable response latency. The findings demonstrate that retrieval-grounded agentic orchestration can significantly improve personalization and adaptability in AI-assisted career preparation systems.

 


Full Text:

PDF

References


A. Gugnani and H. Misra, “Implicit Skills Extraction Using Document Embedding and Its Use in Job Recommendation,” Proceedings of the AAAI Conference on Arti- ficial Intelligence, 2020.

J. Zhao, G. Lan, Z. Huang, Y. Niu, and Q. Liu, “Embedding-Based Recommender System for Job to Candidates Matching,” 2021.

Y.-Z. Lin, C.-H. Chan, T.-Y. Lin, Y.-C. Huang, Y.-C. Chang, J.-Y. Lin, C.-H. Ku, Y.-L. Tsai, W.-C. Yang, and C.-H. Chang, “RAG-PRISM: A Personalized, Rapid, and Immersive Skill Mastery Framework Using Retrieval-Augmented Generation,” Proceedings of ICIT, 2024.

P. Lewis, E. Perez, A. Piktus, et al., “Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks,” Advances in Neural Information Processing Sys- tems (NeurIPS), 2020.

Y. Song, C. Lothritz, D. Tang, T. F. Bissyand´e, and J. Klein, “Revisiting Code Similarity Evaluation with Abstract Syntax Tree Edit Distance,” Proceedings of ACL, 2024.


Refbacks

  • There are currently no refbacks.