Resume Builder Using ATS Score Calculation with Model Context Protocol Servers and Large Language Models

Abstract

Contemporary recruitment processes face significant challenges in efficiently matching candidates with job requirements due to the overwhelming volume of applications and limitations of traditional Applicant Tracking Systems. This paper presents an innovative resume evaluation platform that integrates hybrid Machine Learning models with Large Language Model capabilities through a Model Context Protocol server architecture. The proposed system combines deterministic ATS scoring using TF-IDF vectorization and BERT-based semantic similarity analysis with contextual feedback generation powered by transformer-based LLMs. Implementation utilizes Flask for web application management and FastAPI for MCP server orchestration, enabling real-time resume optimization with transparent scoring mechanisms. Experimental results demonstrate superior performance in resume-job description matching accuracy compared to conventional keyword-based systems, achieving 92% semantic similarity correlation with human recruiter assessments. The platform addresses critical gaps in candidate empowerment, recruiter efficiency, and bias mitigation while maintaining stringent data privacy standards and extensible architecture for future enhancements.