Multi-Agent System for Resume and Job Offer Analysis

Development of a multi-agent system for the automatic analysis of resumes and job offers. The system reads a PDF resume, semantically compares it against a database of 100,000 job offers on Qdrant, evaluates skill compatibility and produces a complete analysis. All agents use Mistral Large 2411 as the LLM and are orchestrated through CrewAI.

The agents: PDF Agent (text extraction and resume synthesis), Mathematical Stats Agent (quantitative metrics and readability), Job Offers Agent (top-5 semantic search on Qdrant), Soft Skills Agent and Hard Skills Agent (1-10 compatibility score per offer), Skill Coherence Evaluator Agent (lexical coherence via Qwen2.5-1.5B + semantic coherence via sentence transformer), Salary Agent (salary range estimation), Summary Agent (final recap) and Chatbot Agent (conversational scenarios).

Three usage scenarios: an automatic sequential pipeline, a single-agent conversational mode via Gradio (one agent, all the tools) and a multi-agent conversational mode, where the Chatbot Agent acts as a manager and delegates requests to the specialized agents.