AI-Driven Consultant Matching : Automating Consultant Selection Using RAG-Enhanced Large Language Models
Gratschew, Max (2025)
2025
Tietotekniikan DI-ohjelma - Master's Programme in Information Technology
Informaatioteknologian ja viestinnän tiedekunta - Faculty of Information Technology and Communication Sciences
Hyväksymispäivämäärä
2025-06-25
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202506247419
https://urn.fi/URN:NBN:fi:tuni-202506247419
Tiivistelmä
Efficient resource allocation plays a critical role in software consultancy, where matching the right consultants to the right projects is essential for success. Despite advancements in resource management tools, the process of identifying available and skilled consultants often remains manual, time-consuming, and prone to inefficiencies. This thesis addresses this challenge by leveraging Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), and automated data integration to improve the process of consultant selection in software consultancy.
The research problem of this thesis was to design and implement a scalable prototype that assists in the consultant selection process by integrating CVs with skillset and certification data from Jira and combining them with project requirements to surface suitable consultant profiles. The proposed system leverages RAG to enhance the capabilities of LLMs, enabling the retrieval of relevant consultant documents and project context to support accurate and context-aware recommendations. While real-time availability integration was considered, it was not implemented due to practical constraints during the study.
The research followed a pragmatic and iterative development approach, combining system design with stakeholder feedback to refine the solution. The implemented tool retrieves skills data and certificate information from Jira Assets, and consultant CVs from Google Drive. This information is processed through a custom backend and integrated with RAG-enabled LLMs to analyze project descriptions and suggest suitable consultant profiles. The tool was evaluated through real-world scenarios and qualitative feedback from team leaders and sales personnel.
The results of this thesis are threefold. Firstly, the system demonstrates how LLMs can reduce the manual effort involved in consultant selection by assisting with early-stage recommendations. Secondly, the integration of RAG improves the contextual relevance of these suggestions by grounding them in retrieved consultant data. Thirdly, the study identifies key challenges and future opportunities in applying RAG and LLMs to resource allocation, particularly around data quality, identity resolution, and real-time integration.
This thesis contributes to the field of data-driven decision-making in software consultancy by demonstrating how emerging AI technologies can be integrated into existing workflows to support resource planning. The findings provide practical guidance for adopting RAG-enhanced LLMs in similar organizational contexts and highlight directions for further development and evaluation.
The research problem of this thesis was to design and implement a scalable prototype that assists in the consultant selection process by integrating CVs with skillset and certification data from Jira and combining them with project requirements to surface suitable consultant profiles. The proposed system leverages RAG to enhance the capabilities of LLMs, enabling the retrieval of relevant consultant documents and project context to support accurate and context-aware recommendations. While real-time availability integration was considered, it was not implemented due to practical constraints during the study.
The research followed a pragmatic and iterative development approach, combining system design with stakeholder feedback to refine the solution. The implemented tool retrieves skills data and certificate information from Jira Assets, and consultant CVs from Google Drive. This information is processed through a custom backend and integrated with RAG-enabled LLMs to analyze project descriptions and suggest suitable consultant profiles. The tool was evaluated through real-world scenarios and qualitative feedback from team leaders and sales personnel.
The results of this thesis are threefold. Firstly, the system demonstrates how LLMs can reduce the manual effort involved in consultant selection by assisting with early-stage recommendations. Secondly, the integration of RAG improves the contextual relevance of these suggestions by grounding them in retrieved consultant data. Thirdly, the study identifies key challenges and future opportunities in applying RAG and LLMs to resource allocation, particularly around data quality, identity resolution, and real-time integration.
This thesis contributes to the field of data-driven decision-making in software consultancy by demonstrating how emerging AI technologies can be integrated into existing workflows to support resource planning. The findings provide practical guidance for adopting RAG-enhanced LLMs in similar organizational contexts and highlight directions for further development and evaluation.