AI-Powered Job Matching System

This project builds a complete pipeline to recommend jobs matching a given resume. It was created for my CS325 final project.

How it Works

1. Data Acquisition:
   • Uses the JSearch API to collect job postings from multiple sources.
   • Focuses on matching jobs to the profile in mock_resume.txt (Jane Doe).
   • Raw data saved to data/jobs_raw.json.
2. Data Preprocessing:
   • Cleans job descriptions and extracts key fields.
   • Saves to data/jobs_cleaned.json.
3. Embedding Information:
   • Uses sentence-transformers (all-MiniLM-L6-v2) to embed:
     • Each job posting → embeddings/jobs_embeddings.npy
     • The resume → embeddings/resume_embedding.npy
4. Similarity Calculation:
   • Computes cosine similarity between the resume and each job.
   • Outputs the top 10 matches to results/top_matches.json.

File Overview

• scripts/scrape_jobs.py → Gets job data from JSearch API.
• scripts/preprocess.py → Cleans and structures the data.
• scripts/embed_resume.py → Generates embedding for the resume.
• scripts/embed_jobs.py → Computes cosine similarity and saves top matches.
• .env → Stores sensitive API keys (excluded from Git).

Output Files

• data/jobs_cleaned.json → Preprocessed job data.
• embeddings/ → Contains .npy embedding files for jobs & resume.
• results/top_matches.json → JSON listing top 10 job matches.

How to Run

pip install -r requirements.txt
# or at a minimum
pip install requests sentence-transformers numpy python-dotenv

Then create a .env file in your project root with your keys: RAPIDAPI_KEY = your_rapidapi_key_here HF_API_TOKEN = your_huggingface_token_here

Then run each stage: python scripts/scrape_jobs.py python scripts/preprocess.py python scripts/embed_resume.py python scripts/embed_jobs.py python scripts/match_jobs.py