Better ATS Service

A FastAPI-based microservice that processes resumes against job requirements using AI. The service extracts candidate information from uploaded documents and assesses each job requirement against the candidate's experience.

Related blog post

Thought exercise: Improving Candidate Experience in ATS with AI-Powered Resume Parsing

Features

• Extract structured candidate data from resumes (PDF format)
• Assess job requirements against candidate experience
• Generate clarifying questions for requirements that can't be verified
• Parallel processing of requirements
• RESTful API interface

Project Structure

better-ats-service/
├── models/              # Pydantic models for data validation
│   ├── experience.py    # Experience data model
│   ├── candidate.py     # Candidate data model
│   ├── assessment.py    # Requirement assessment model
│   └── process_input.py # API input/output models
├── routes/              # FastAPI route handlers
│   └── process.py       # Main processing endpoint
├── pipelines/           # Haystack pipeline definitions
│   ├── assessment_pipeline.py    # Job requirement assessment
│   └── candidate_data_pipeline.py # Candidate data extraction
├── exec/                # Pipeline execution modules
│   ├── exec_assessment.py     # Parallel requirement processing
│   └── exec_candidate_data.py # Candidate data extraction
└── main.py             # FastAPI application entry point

Environment Variables

Create a .env file in the root directory with the following variables:

# Server Configuration
HOST=0.0.0.0            # Server host
PORT=8000               # Server port

# OpenAI Configuration
OPENAI_API_KEY=your_key # Your OpenAI API key

# Model Configuration
CANDIDATE_DATA_MODEL=gpt-4  # Model for candidate data extraction
ASSESSMENT_MODEL=gpt-4      # Model for requirement assessment

# Performance Configuration
MAX_WORKERS=4           # Maximum concurrent requirement assessments

# File Storage Configuration
UPLOAD_TMP_DIR=/tmp    # Temporary directory for file uploads

API Interface

POST /process

Process resumes against job requirements.

Request

curl -X POST http://localhost:8000/process \
  -H "Content-Type: multipart/form-data" \
  -F "files=@/path/to/resume1.pdf" \
  -F "files=@/path/to/resume2.pdf" \
  -F 'process_input={
    "job_requirements": [
      "5+ years of Python development experience",
      "Experience with AWS cloud services",
      "Strong background in machine learning"
    ]
  }'

Response

{
  "candidate_data": {
    "first_name": "John",
    "last_name": "Doe",
    "email": "john.doe@email.com",
    "phone": "+1 (555) 123-4567",
    "linkedin": "linkedin.com/in/johndoe",
    "experiences": [
      {
        "company": "Tech Corp",
        "title": "Senior Python Developer",
        "start_date": "01/2020",
        "end_date": "Present",
        "description": "Led development of cloud-based applications..."
      }
    ]
  },
  "requirements_assessment": [
    {
      "requirement": "5+ years of Python development experience",
      "present_in_documents": true,
      "inquiry": null
    },
    {
      "requirement": "Experience with AWS cloud services",
      "present_in_documents": true,
      "inquiry": null
    },
    {
      "requirement": "Strong background in machine learning",
      "present_in_documents": false,
      "inquiry": "Could you describe any specific machine learning projects you've worked on?"
    }
  ]
}

Installation

1. Clone the repository:

git clone https://github.com/yourusername/better-ats-service.git
cd better-ats-service

2. Create and activate a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

4. Set up environment variables:

cp .env.example .env
# Edit .env with your configuration

5. Run the service:

uvicorn main:app --reload

The API documentation will be available at http://localhost:8000/docs

Dependencies

• FastAPI: Web framework
• Haystack: Document processing and LLM pipelines
• PyPDF: PDF processing
• OpenAI: LLM provider
• python-multipart: File upload handling
• python-dotenv: Environment variable management

Development

The service uses several key components:

• Models: Pydantic models for data validation and serialization
• Routes: FastAPI route handlers for API endpoints
• Pipelines: #Haystack pipelines for document processing and LLM interactions
• Exec: Execution modules for parallel processing and pipeline orchestration

Adding New Features

1. Define new models in models/
2. Create new pipelines in pipelines/
3. Add execution logic in exec/
4. Define routes in routes/

License

This project is licensed under the MIT License - see the LICENSE.md file for details.