🚀 Project Overview
The Skill Match project is an advanced, hybrid solution designed to accurately extract, standardize, and validate latent professional skills from unstructured documents (like PDF resumes) for use in downstream job-matching and recommendation systems.
We overcome the limitations of traditional keyword parsers and single LLM approaches by orchestrating a multi-stage workflow built on two core scientific components: Google's Gemini Chat Model for initial semantic extraction and final validation, and a custom Transformer service for systematic data scoring and enrichment.
🛠️ Architecture and Components
The system is deployed using a modular, event-driven architecture:
Frontend (Loveable): A user interface for uploading resumes, triggering the workflow via a webhook.
Orchestration (n8n): Manages the 7-node pipeline, handling PDF conversion, data flow control, and API calls.
Data Processing:
Gemini (Extraction & Validation): Used to generate structured JSON from raw text and later validate and map skills to a recognized taxonomy (e.g., ESCO).
Transformer (Enrichment): An external service that systematically processes candidate skills (moving beyond raw LLM output) to refine, score, and embed them based on proprietary research. (refer to https://doi.org/10.1016/j.simpa.2025.100772 and https://github.com/KonstantinosPetrakis/esco-skill-extractor)