HR Matchmaker

Overview

The HR Matchmaker is an advanced application that automates recruitment screening by matching job applicants with positions using state-of-the-art Natural Language Processing (NLP) and machine learning techniques. The system leverages transformer models and optimized processing for Apple Silicon, providing real-time matching and interactive visualization.

System Architecture

1. Machine Learning Core

• Dual Model Support
  • Transformer-based Model (Default)
    • Uses SentenceTransformer with MiniLM architecture
    • Optimized for Apple Silicon (M1/M2) using MPS
    • Automatic fallback to CPU when needed
  • TF-IDF Model (Alternative)
    • N-gram analysis (up to trigrams)
    • Custom vocabulary size and document frequency thresholds
    • Sublinear scaling for better term weighting

2. Data Processing Pipeline

• Input Processing

  • Automated CSV file monitoring
  • Real-time data validation
  • Text preprocessing and standardization

• Model Training

  • Dynamic model selection via environment variables
  • Automatic retraining on data updates
  • Optimized batch processing
  • Hardware-specific optimizations

• Matching Engine

  • Cosine similarity computation
  • Configurable matching thresholds
  • Batch-processed scoring
  • Memory-efficient operations

3. Interactive Dashboard

• Control Panel

  • Job position selection
  • Minimum match score adjustment
  • Maximum applicant display limit
  • Real-time job requirement display

• Visualization

  • Interactive match score charts
  • Detailed applicant profiles
  • Expandable resume sections
  • Score-based ranking

Technical Features

Machine Learning

• Transformer Model

  • Model: all-MiniLM-L6-v2
  • Hardware acceleration support
  • Batch processing optimization
  • Automatic device selection

• TF-IDF Model

  • N-gram range: (1, 3)
  • Max features: 15,000
  • Frequency thresholds: min_df=2, max_df=0.85
  • Sublinear TF scaling

Data Management

• File Monitoring

  • Real-time CSV change detection
  • Automatic model retraining
  • Error handling and logging

• Data Validation

  • Schema validation
  • Data completeness checks
  • Duplicate detection
  • Error reporting

Installation

1. Environment Setup

   # Create Python 3.11 virtual environment
   python3.11 -m venv thevenv
   source thevenv/bin/activate  # or `thevenv\Scripts\activate` on Windows

2. Install Dependencies

   pip install -r requirements.txt

Usage

1. Start the Application

   # Set model type (optional)
   export MODEL_TYPE=transformer  # or 'tfidf'
   
   # Run the main application
   python src/main.py

2. Launch Dashboard

   streamlit run src/dashboard.py

Configuration

Model Selection

• Set MODEL_TYPE environment variable:
  • transformer: Use transformer-based model (default)
  • tfidf: Use TF-IDF based model

Hardware Optimization

• Automatic detection of Apple Silicon
• MPS acceleration when available
• Graceful fallback to CPU processing

System Components

Core Files

• transformer_model.py: Transformer-based matching implementation
• model_training.py: Model training and evaluation logic
• data_preprocessing.py: Data cleaning and preparation
• applicant_screening.py: Candidate matching logic
• dashboard.py: Interactive UI implementation
• file_monitor.py: Data change monitoring
• validation.py: Data validation
• config.yaml: System configuration
• logger.py: Logging system

Data Files

• job_posts.csv: Job requirements and descriptions
• applicants.csv: Applicant profiles and resumes
• screening.log: System activity logs

Performance Considerations

• Batch processing for large datasets
• Hardware-specific optimizations
• Memory-efficient operations
• Automatic model versioning

Future Enhancements

1. ML Improvements

• Additional transformer models
• Custom model fine-tuning
• Multi-lingual support

2. System Features

• API integration
• Automated report generation
• Advanced analytics dashboard

3. Performance

• Distributed processing
• Advanced caching
• Real-time updates

Maintenance

• Monitor logs/screening.log
• Regular model performance evaluation
• Data quality checks
• System health monitoring

Support

For technical issues:

• Check system logs
• Verify data format
• Review configuration
• Contact system administrator

Contributing

1. Fork repository
2. Create feature branch
3. Implement changes
4. Submit pull request

License

MIT License