---

🧠 AI-Powered CV to Job Matcher

This project is an AI-based web scraping and job matching tool that leverages OCR, NLP, and LLMs to read a user's CV and suggest job opportunities that align with the candidate’s skills and experience.

---

🚀 Features

• 📝 CV Parsing: Converts PDF CVs into structured text using OCR and text block analysis.
• 🧠 Semantic Matching: Embeds CV content into a vector space for similarity-based job relevance matching using Weaviate.
• 🌐 Web Scraping: Scrapes job listings using crawl4ai and extracts structured data using LLM-based extraction.
• 📄 Output: Saves top matched jobs in a CSV for easy viewing and integration.

---

📁 Project Structure

├── ocr.py                # Converts PDF to structured text
├── main.py               # Embeds CV into Weaviate, runs scraper and LLM matching
├──  venue.py             # Contains scraping and job extraction logic
├── models/
│   └── venue.py          # Job schema model
├── utils/
│   ├── data_utils.py     # Helpers for saving data and filtering
│   └── scraper_utils.py  # Scraper-related utilities
├── extracted_texts/      # Output folder for OCR-processed text
├── extracted_images/     # Extracted images from the CV (if any)
├── complete_venues.csv   # Final matched job listings

---

⚙️ Technologies Used & Why

🔹 Python

Used for overall backend scripting due to its rich ecosystem in AI, OCR, and web scraping.

🔹 PyMuPDF (fitz)

Used to extract both text and images from the uploaded PDF CV. It provides precise control over layout blocks, fonts, and text formatting.

🔹 Tesseract OCR (pytesseract)

Used to convert any images (scanned signatures, certificates, etc.) inside the CV to text. Helps enhance the completeness of the CV content before analysis.

🔹 LangChain

Used to split the CV into logical chunks and interface with vector databases like Weaviate. It simplifies working with large documents and embeddings.

🔹 HuggingFace Transformers

Provides high-quality language embeddings that represent the semantic meaning of the CV content. These are then stored in Weaviate.

🔹 Weaviate (Vector Database)

Used to store the semantic vector representation of the CV chunks and perform similarity search against job prompts. Weaviate is fast, scalable, and integrates smoothly with LangChain.

🔹 crawl4ai

An intelligent web scraping framework powered by LLMs. It allows automatic extraction of structured data from HTML pages using prompt-driven extraction logic.

🔹 AsyncIO

Used to perform concurrent scraping across multiple pages without blocking, improving speed and performance.

🔹 LLM Extraction Strategy (groq/deepseek)

Used to extract only the most relevant jobs based on the user’s CV context. It ensures meaningful matches instead of random job listings.

---

🧪 How It Works

1. Step 1 – Add your CV
   Place your cv.pdf inside the root directory of the project.

2. Step 2 – Run OCR
   Execute the following to extract all text and images from the PDF:

   python ocr.py

   • 📂 Output: extracted_texts/cv.txt (structured text of your CV)

3. Step 3 – Run Job Matching
   Run the main logic that matches your CV content to current job listings:

   python main.py

   This script:

   • Embeds your CV
   • Uploads chunks into Weaviate
   • Scrapes job listings
   • Matches them using LLM extraction logic
   • Saves final job matches in complete_venues.csv

---

🧩 Internal Workflow

cv.pdf
  ↓ (ocr.py)
cv.txt
  ↓ (main.py)                     ↓ (scrape.py)
Vectorized Embeddings → Weaviate ← Scraped Jobs ← Dialog.lk
         ↓                                ↓
     LLM-based Matching (LLMExtractionStrategy)
         ↓
     Output CSV: complete_venues.csv

---

✅ Requirements

• Python 3.8+
• Install dependencies:

  pip install -r requirements.txt

---

📌 Notes

• Ensure Tesseract-OCR is installed on your machine and its path is correctly set in ocr.py.
• Make sure to update your .env file with the GROQ_API_KEY and any other necessary keys.

---

🧠 Future Improvements

• Upload CVs via a simple web interface
• Add support for multiple LLM providers (e.g., OpenAI, Anthropic)
• Improve job source diversity (LinkedIn, Glassdoor, etc.)

---

👨‍💻 Author

Made with ❤️ by Gihan Lakmal