Job Ranker

🚀 AI-powered job ranking using local NLP and embeddings

Overview

Job Ranker is a local-first, AI-powered job aggregation and ranking system designed to optimize the job search process.

It collects job offers from multiple platforms, extracts required skills using a small language model, and ranks job postings based on their relevance to a candidate profile using multilingual embeddings.

The project was born from a real-world problem: job searching is time-consuming and often involves browsing through many irrelevant offers. Job Ranker aims to reduce this friction by surfacing the most relevant opportunities first.

---

Key Features

• 🔎 Multi-source job scraping (LinkedIn, Indeed – more to come)
• 🧠 Skill extraction using a Small Language Model (SLM)
• 📐 Semantic job ranking using embeddings and cosine similarity
• ⚡ Fully local execution (no GPU required)
• 🗂️ Centralized job storage with SQLite
• 🌐 Web interface for browsing and ranking job offers

---

Architecture (High Level)

1. Data Collection

   • Job scraping from multiple platforms using jobspy
   • Data stored in a Pandas DataFrame and persisted in a SQLite database

2. AI Processing & Ranking

   • Skill extraction from job descriptions using Gemma 3 (1B)

     • Chosen for its multilingual capabilities and efficiency on CPU

   • Skill representation using bge-m3 embeddings

     • Lightweight, multilingual, and among the best-performing embedding models

   • Job relevance computed via cosine similarity

   • Offers are ranked according to their semantic alignment with the candidate profile

3. Visualization

   • Web-based interface to explore, filter, and prioritize job offers

---

Tech Stack

• Language: Python

• Scraping: jobspy

• NLP / AI:

  • Gemma 3 (1B) – skill extraction
  • bge-m3 – multilingual embeddings

• Data: Pandas, SQLite

• Similarity: Cosine similarity

• Web: Web interface documentation

---

Requirements

• Python 3.11+
• CPU-only machine (tested on 8 cores / 16 GB RAM)
• No GPU required
• Ollama (Download here)

---

Installation

1. Python dependencies

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

2. Install Ollama and pull the model

# Install Ollama from https://ollama.ai/download
# Then pull the required model:
ollama pull gemma3:1b

---

Usage

Run full pipeline (scraping + ranking)

python script.py -c example_config.yaml

Run ranking only (skip scraping)

python script.py -c example_config.yaml --process-only

The configuration file allows you to define:

• Job sources
• Search keywords
• Target skills
• Language preferences

---

Project Status & Roadmap

🚧 Work in progress

Planned improvements:

• Support for additional job platforms

If you are reading this, feel free to contribute.

---

Motivation

Job Ranker was developed alongside my own job search as a Data Scientist.

Beyond its practical use, the project serves as a real-world application of:

• NLP for unstructured text
• Efficient AI pipelines under hardware constraints
• End-to-end data and AI system design

---

License

This project is licensed under the GNU General Public License v3.0 (GPLv3).