HireIndex: A RAG-Enhanced AI Recruitment Platform for Fair, Transparent, and Efficient Hiring

Automating resume screening while ensuring fairness and transparency remains a major challenge for modern recruitment systems. We present HireIndex, an end-to-end recruitment pipeline that integrates a lightweight transformer model (DistilBERT) for semantic resume–job matching with a Retrieval-Augmented Generation (RAG) layer to incorporate domain-specific knowledge at inference time. HireIndex includes bias detection and mitigation modules (adversarial debiasing and post-processing constraints) and produces recruiter-facing expla nations through SHAP-based feature attributions and naturallanguage rationales. Evaluated on a curated dataset of 10,000 resume–job pairs and a held-out test set of 223 real resumes, the system demonstrates promising accuracy (98% on the heldout test set) while reducing average human review time and improving fairness metrics. Although based on a relatively small and domain-specific dataset, HireIndex primarily contributes as a system-level integration of existing techniques rather than a novel algorithmic framework. We detail its architecture, training workflow, fairness evaluations, and ethical safeguards.