Synapse: Evolving Job-Person Fit with Explainable Two-phase Retrieval and LLM-guided Genetic Resume Optimization

TL;DR

Synapse is a multi-stage, explainable recruitment system combining dense retrieval, LLM reasoning, and evolutionary resume optimization to improve candidate-job matching at scale.

Contribution

It introduces a novel multi-stage retrieval framework with an explainability layer and a black-box resume optimization method guided by LLMs, enhancing matching accuracy.

Findings

Ensemble improves nDCG@10 by 22% over baseline.

Evolutionary optimization yields over 60% relative gain in profile scoring.

System demonstrates monotonic improvements in candidate-job alignment.

Abstract

Modern recruitment platforms operate under severe information imbalance: job seekers must search over massive, rapidly changing collections of postings, while employers are overwhelmed by high-volume, low-relevance applicant pools. Existing recruitment recommender systems typically rely on keyword matching or single-stage semantic retrieval, which struggle to capture fine-grained alignment between candidate experience and job requirements under real-world scale and cost constraints. We present Synapse, a multi-stage semantic recruitment system that separates high-recall candidate generation from high-precision semantic reranking, combining efficient dense retrieval using FAISS with an ensemble of contrastive learning and Large Language Model (LLM) reasoning. To improve transparency, Synapse incorporates a retrieval-augmented explanation layer that grounds recommendations in explicit evidence. Beyond retrieval, we introduce a novel evolutionary resume optimization framework that treats resume refinement as a black-box optimization problem. Using Differential Evolution with LLM-guided mutation operators, the system iteratively modifies candidate representations to improve alignment with screening objectives, without any labeled data. Evaluation shows that the proposed ensemble improves nDCG@10 by 22% over embedding-only retrieval baselines, while the evolutionary optimization loop consistently yields monotonic improvements in recommender scores, exceeding 60% relative gain across evaluated profiles. We plan to release code and data upon publication.