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

A research team has introduced Synapse, a novel AI system designed to tackle the information imbalance in modern recruitment. The system employs a two-phase approach: first, it uses efficient dense retrieval via FAISS to generate a broad pool of candidates (high recall), then applies a semantic reranking ensemble combining contrastive learning and Large Language Model (LLM) reasoning for high precision. This architecture addresses the limitations of traditional keyword-matching or single-stage systems, capturing fine-grained alignment between candidate experience and job requirements. Crucially, Synapse includes a retrieval-augmented explanation layer, grounding its recommendations in explicit evidence to improve transparency for both job seekers and employers.

Beyond intelligent matching, Synapse pioneers a novel application of evolutionary algorithms to resume optimization. It frames resume refinement as a black-box optimization problem, using a technique called Differential Evolution. Here, LLMs act as intelligent mutation operators, iteratively suggesting modifications to a candidate's resume representation to better align with target job descriptions. This process requires no labeled training data and operates autonomously. In evaluations, the full Synapse ensemble demonstrated a 22% improvement in the nDCG@10 metric over embedding-only baselines. Furthermore, the evolutionary optimization loop consistently produced monotonic improvements, yielding over a 60% relative gain in recommender scores across the tested candidate profiles. The team plans to release the code and data upon publication.