Bridging the Enterprise Skills Gap: A Multi-Modal Agentic RAG Architecture for Autonomous Talent Discovery
DOI:
https://doi.org/10.63282/3050-9262.IJAIDSML-V7I2P110Keywords:
Agentic RAG, Multi-Modal Generative AI, Azure AI Search, Human Capital Management (HCM), Enterprise Architecture, Semantic Search, Document Intelligence, Vector Databases, HyDE, Sidecar PatternAbstract
In the modern enterprise, Human Capital Management (HCM) suffers from a systemic "data-rich, insight-poor" paradox. While platforms like Workday house vast repositories of structured metadata, the most critical indicators of talent niche technical competencies, complex project histories, and latent expertise remain trapped in unstructured formats such as PDF resumes and free-text fields. This paper introduces "Colleague Search," a novel Agentic Multi-Modal Retrieval-Augmented Generation (RAG) framework designed to transform static HR databases into an autonomous discovery engine. By utilizing Azure AI Search, Azure Document Intelligence, and a proprietary "Sidecar Metadata Pattern," this architecture moves beyond simple keyword retrieval to achieve high-fidelity visual document intelligence and semantic reasoning. The implementation effectively solves the "Dark Data" problem inherent in complex resume tables and legacy ingestion pipelines, reducing the talent discovery cycle from 72 hours to sub-second, high-precision identification. Achieving a 98% accuracy rate in skill extraction, this framework serves as a scalable blueprint for autonomous internal mobility, demonstrating how orchestrated AI can fundamentally reshape enterprise resource planning.
References
[1] Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., ... & Stoyanov, V. (2020). "Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks." Advances in Neural Information Processing Systems (NeurIPS), 33, 9459-9474.
[2] Liu, N. F., Lin, K., Hewitt, J., Paranjape, A., Bevilacqua, M., Petroni, F., & Liang, P. (2023). "Lost in the Middle: How Language Models Use Long Contexts." Transactions of the Association for Computational Linguistics.
[3] Gao, L., Dai, Z., Pasupat, P., Chen, A., Kumar, A., & Callan, J. (2022). "Precise Zero-Shot Dense Retrieval without Relevance Labels." Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (ACL).
[4] Yao, S., Zhao, J., Yu, D., Du, N., Shafran, I., Narasimhan, K., & Cao, Y. (2022). "ReAct: Synergizing Reasoning and Acting in Language Models." International Conference on Learning Representations (ICLR).
[5] Yan, S., Gu, J., Zhu, Y., & Qiu, X. (2024). "Corrective Retrieval Augmented Generation." arXiv preprint arXiv:2401.15884.
[6] Smock, D., Pesala, R., & Borthwick, R. (2022). "PubTables-1M: Towards comprehensive table extraction from unstructured documents." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).
[7] Microsoft Corporation. (2024). "Azure AI Search: Hybrid Search and Semantic Ranking Architecture." Azure Architecture Center Technical Documentation. Retrieved from Microsoft Learn.
