Explainable Resume to Job Matching and Shortlisting Agent

Can a small local language model screen a resume for a job and explain itself? This project builds a Microsoft AutoGen team: a screener labels the fit from rule-based skill lists and redacted excerpts, and an auditor runs written rule checks and sends problems back. The model (qwen2.5:3b) runs in local Ollama on a laptop with RTX 2050 4 GB (Ryzen 7, 8 GB RAM).

NOTE: Research and teaching demo, not a hiring tool. A person must decide; the agent only suggests and sends nothing to candidates. Employment screening is high-risk under the EU AI Act and is regulated in many places. On this benchmark the agent scored below a simple classical model, and the labels are noisy. Fairness was only probed with fictional name and pronoun cues on benchmark resumes, and that does not show the system is fair.

Category: Department of Computer Science and Engineering / Agent Development

Agent Development Project AG19 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.