Job Recommendation System

Machine learning-based job recommendation system that classifies job postings and generates personalized recommendations using NLP and ML.

Features

• Job Domain Classification: Automatically classify job postings into domains
• Skill Relevancy Scoring: Compute skill and text similarity between candidates and jobs
• Recommendation Engine: Generate personalized job recommendations
• REST API: FastAPI-based endpoints with OpenAPI docs

Project Structure

src/
  api/
    main.py                # FastAPI app entrypoint
    models/schemas.py      # Pydantic request/response models
    routes/
      classification.py    # /classification endpoints
      recommendations.py   # /recommendations endpoints
  data_processing/
    csv_processor.py       # CSV loaders/cleaners and utilities
  job_domain_classifier/
    classifier.py          # Training, inference, model IO
    preprocessing.py       # Text cleaning and domain assignment
  matching_engine/
    candidate_job_matcher.py
    recommendation_generator.py
  resume_parser/
    llama_parser.py
  skill_relevancy_scorer/
    fuzzy_matcher.py
    relevancy_calculator.py
    skill_extractor.py
  utils/
    file_utils.py
    ml_utils.py

config/
  settings.py              # Paths, API config, env vars
  logging_config.py        # Logging setup

scripts/
  process_data.py          # Organize/process CSV data
  train_models.py          # Train and save classifier

data/
  raw/                     # Place input CSVs (e.g., job_details.csv)
  processed/               # Generated processed datasets
  models/                  # Saved models

tests/                     # Unit tests

Requirements

• Python 3.8+

Installation

Install in editable mode using the included packaging metadata:

pip install -e .

This project uses configuration from environment variables (optional). Create a .env in the project root to override defaults:

API_HOST=127.0.0.1
API_PORT=8000
MLFLOW_TRACKING_URI=.\mlflow

Data Setup

Place the required CSV files into data/raw/:

• job_details.csv (required for recommendations and training)
• Optional: all_resumes_data.csv, job_details_with_predictions.csv, etc.

You can also run the helper script to organize/process data:

python scripts/process_data.py

Running the API

Start the FastAPI server:

uvicorn src.api.main:app --reload --host 127.0.0.1 --port 8000

Open the interactive docs at http://127.0.0.1:8000/docs or visit /docs when running the server.

API Overview

• Note: The classification endpoint requires a trained model saved under data/models/. Run training first if you haven't.
• GET / – Root service info
• GET /recommendations/health – Health check
• POST /classification/predict – Predict job domain
  • Request body:

    { "description": "We are looking for a Python developer with ML experience" }

  • Example response:

    { "domain": "Software Development", "confidence": 0.8, "cleaned_text": "python developer ml experience" }

• POST /recommendations/generate – Generate personalized job recommendations
  • Request body:

    {
      "candidate": {
        "name": "Jane Doe",
        "email": "[email protected]",
        "skills": ["python", "machine learning", "sql"],
        "experiences": "Worked on data pipelines",
        "education": "B.Tech Computer Science",
        "domains": ["Data Science", "Software Development"]
      },
      "max_recommendations": 5,
      "filters": {"location": "Remote"}
    }

  • Example response (array):

    [
      {"job_id": "123", "title": "Python Developer", "company": "Tech Corp", "location": "Remote", "relevancy_score": 0.92}
    ]

Training the Classifier

Ensure data/raw/job_details.csv exists, then run:

python scripts/train_models.py

The best model will be saved under data/models/.

Running Tests

python -m unittest discover -s tests

Docker

Build the image and run the API:

docker build -t job-reco:dev .
docker run --rm -p 8000:8000 -e API_HOST=0.0.0.0 -e API_PORT=8000 -v %cd%/data:/app/data job-reco:dev

Using docker-compose (recommended for development):

docker compose up --build

What this does:

• Builds from Dockerfile
• Serves the API on port 8000 with hot reload
• Mounts ./src, ./config, ./scripts, and ./data into the container

Environment variables (can be set via a .env file at project root):

• API_HOST (default 0.0.0.0)
• API_PORT (default 8000)
• MLFLOW_TRACKING_URI (optional)

Optional database:

• Not required by default. A Postgres service is stubbed and commented in docker-compose.yml. Uncomment and configure if you add persistence later.

Development Notes

• Use pip install -e . to ensure imports like src.api resolve correctly during development.
• Default directories are created automatically (data/raw, data/processed, data/models).
• Logging is configured via config/logging_config.py and used across scripts and API.