🎯 AI-Powered CV Analysis System

Advanced CV screening and candidate evaluation using Qwen2.5:1.5B-Instruct AI model

An intelligent recruitment assistant that extracts, analyzes, and ranks CV candidates using state-of-the-art AI technology. This system provides comprehensive candidate evaluation with anti-hallucination measures to ensure accurate and reliable results.

🎯 Overview

This CV Analysis System leverages the Qwen2.5:1.5B-Instruct language model to provide intelligent, automated CV screening for recruitment processes. The system is designed to:

Extract structured information from PDF CVs with high accuracy
Generate professional candidate summaries using AI-powered natural language processing
Provide detailed reasoning for job fit assessment with transparent scoring
Rank candidates based on customizable job requirements
Prevent AI hallucination through strict prompt engineering and validation

🎯 Target Use Cases

HR Departments: Automate initial CV screening for large candidate pools
Recruitment Agencies: Provide consistent, objective candidate evaluation
Hiring Managers: Get detailed insights and reasoning for hiring decisions
Small Businesses: Access enterprise-level CV analysis without complex setup

✨ Key Features

🤖 AI-Powered Analysis

Qwen2.5:1.5B-Instruct Model: State-of-the-art language model for intelligent text understanding
Contextual Understanding: Goes beyond keyword matching to understand candidate qualifications
Natural Language Reasoning: Provides human-readable explanations for fit assessment

📄 Document Processing

PDF Text Extraction: Robust extraction from various PDF formats
Multi-language Support: Handles CVs in different languages (model dependent)
Error Handling: Graceful handling of corrupted or unreadable files

🎯 Job Matching System

Customizable Job Templates: Pre-built templates for common roles
Interactive Job Creation: Step-by-step job requirement setup
Flexible Scoring: Weighted scoring based on skills, experience, and education

🛡️ Anti-Hallucination Technology

Strict Prompt Engineering: Prevents AI from adding false information
Fact-Based Analysis: Only considers explicitly mentioned qualifications
Conservative Scoring: Ensures reliable and trustworthy results

📊 Comprehensive Reporting

Detailed Candidate Summaries: Professional summaries highlighting key strengths
Fit Score Analysis: Numerical scoring with detailed reasoning
Batch Processing: Analyze multiple CVs simultaneously with ranking
JSON Export: Machine-readable results for integration with other systems

🏗️ System Architecture

📁 chatbot-sso/
├── 📄 main.py                          # Main application entry point
├── 📄 requirements.txt                 # Python dependencies
├── 📁 analyzer/                        # Core analysis modules
│   ├── 📄 cv_analyzer.py              # Alternative CLI interface
│   └── 📁 processors/                 # Processing engines
│       └── 📄 simple_cv_analyzer.py   # Qwen2.5 CV analysis engine
├── 📁 cv/                             # Sample CV storage
├── 📁 cv_test_1/                      # Test CV files
├── 📄 cv_analysis_*.json              # Generated analysis results
└── 📁 venv/                          # Python virtual environment

🔄 Processing Pipeline

graph TD
    A[PDF Input] --> B[Text Extraction]
    B --> C[AI Information Extraction]
    C --> D[Candidate Summary Generation]
    D --> E[Job Fit Analysis]
    E --> F[Scoring & Ranking]
    F --> G[JSON Output]
    
    H[Job Requirements] --> E
    I[Anti-Hallucination Rules] --> C
    I --> D
    I --> E

🚀 Quick Start Guide

📋 Prerequisites

Python 3.8+ (Recommended: Python 3.9 or 3.10)
4GB+ RAM (8GB+ recommended for optimal performance)
3GB disk space for AI model download
Internet connection for initial model download

⚡ Installation

Clone or download the project

git clone <repository-url>
cd cv-analysis-system

Create virtual environment (Recommended)

python -m venv venv

# Windows
venv\Scripts\activate

# macOS/Linux
source venv/bin/activate

Install dependencies

pip install -r requirements.txt

Verify installation

python main.py

🎬 First Run

Run the application

python main.py

Follow the interactive setup:
- Choose a job template or create custom requirements
- Select CV files to analyze
- Review results
Check generated files:
- Look for cv_analysis_*.json files in the project directory

📖 Detailed Usage

🎯 Interactive Mode (Recommended)

The main application provides a guided experience:

python main.py

Step-by-Step Process:

Job Description Setup
- Choose from predefined templates (Data Analyst, Software Engineer, Project Manager)
- Create custom job requirements interactively
- Load job description from JSON file
CV Source Selection
- Analyze single CV file
- Process entire folder of CVs
- Use default CV folder (../cv)
AI Model Initialization
- Automatic download and loading of Qwen2.5 model
- System optimization based on available resources
Analysis Execution
- Real-time progress updates
- Error handling and reporting
- Automatic result saving

🖥️ Command Line Mode

For automation and scripting:

# Direct analyzer usage
cd analyzer
python cv_analyzer.py --file ../cv/resume.pdf --detailed
python cv_analyzer.py --folder ../cv --output results.json

📝 Job Template Examples

Data Analyst Template

{
    'title': 'Senior Data Analyst',
    'department': 'Analytics & Business Intelligence',
    'requirements': {
        'education': 'Bachelor degree in Statistics, Mathematics, Computer Science, or related field',
        'skills': ['Python', 'SQL', 'Data Analysis', 'Excel', 'Tableau', 'Power BI', 'Statistics'],
        'experience_years': 3,
        'domain_knowledge': ['Business Intelligence', 'Data Visualization', 'Statistical Analysis'],
        'soft_skills': ['Communication', 'Problem Solving', 'Team Collaboration'],
        'preferred': ['Machine Learning', 'R', 'Advanced Excel', 'Database Management']
    }
}

Software Engineer Template

{
    'title': 'Software Engineer',
    'department': 'Engineering',
    'requirements': {
        'education': 'Bachelor degree in Computer Science, Software Engineering, or related field',
        'skills': ['Python', 'JavaScript', 'SQL', 'Git', 'REST APIs', 'Database Design'],
        'experience_years': 2,
        'domain_knowledge': ['Web Development', 'Software Architecture', 'Database Management'],
        'soft_skills': ['Problem Solving', 'Team Collaboration', 'Communication'],
        'preferred': ['React', 'Node.js', 'Docker', 'AWS', 'CI/CD']
    }
}

⚙️ Configuration

🎛️ Job Requirements Configuration

You can modify job requirements in several ways:

Interactive Creation (main.py)
Direct Code Modification (analyzer/processors/simple_cv_analyzer.py)
JSON File Loading

Custom Job JSON Format

{
  "title": "Your Job Title",
  "department": "Department Name",
  "requirements": {
    "education": "Education requirements",
    "skills": ["Skill1", "Skill2", "Skill3"],
    "experience_years": 3,
    "domain_knowledge": ["Domain1", "Domain2"],
    "soft_skills": ["Communication", "Leadership"],
    "preferred": ["Nice-to-have1", "Nice-to-have2"]
  }
}

🔧 AI Model Configuration

The system uses Qwen2.5:1.5B-Instruct with optimized parameters:

# Generation parameters (in simple_cv_analyzer.py)
temperature=0.3,        # Conservative for accuracy
top_p=0.6,             # Focused sampling
repetition_penalty=1.1  # Reduce repetition

💾 Memory Optimization

For systems with limited memory:

Reduce batch size in code
Use CPU instead of GPU if memory issues occur
Close other applications during processing

📊 Output Examples

📱 Console Output

🎯 CV ANALYZER - AI-Powered Recruitment Assistant
==================================================

📋 JOB DESCRIPTION SETUP
------------------------
Selected: Senior Data Analyst

🤖 Initializing AI CV Analyzer...
✅ CV Analyzer initialized successfully!

📁 CV SELECTION
---------------
✅ Found 5 PDF files in default folder

🚀 Starting CV Analysis...
--------------------------

🏆 TOP CANDIDATES (by fit score):
 1. john_smith_cv.pdf                   87% (Yes)
 2. sarah_jones_resume.pdf              73% (Yes)
 3. mike_wilson_cv.pdf                  65% (Yes)
 4. lisa_brown_resume.pdf               52% (Maybe)
 5. tom_davis_cv.pdf                    38% (No)

💾 Results saved to: cv_analysis_batch_20250829_130423.json

📄 JSON Output Structure

{
  "analysis_timestamp": "2025-08-29T13:04:23.456789",
  "total_analyzed": 5,
  "successful": 5,
  "results": [
    {
      "filename": "john_smith_cv.pdf",
      "file_path": "/path/to/cv/john_smith_cv.pdf",
      "status": "success",
      "extracted_info": "**Full Name:** John Smith\n**Contact:** john.smith@email.com, +1234567890\n**Education:** Bachelor of Computer Science, MIT (2018)\n**Technical Skills:** Python, SQL, Tableau, Power BI, Statistics, Machine Learning\n**Experience:** 5 years in data analysis roles\n**Summary:** Experienced data analyst with strong technical background",
      "candidate_summary": "John Smith is an experienced data analyst with 5 years of professional experience and a strong educational background from MIT. His technical expertise includes Python, SQL, Tableau, Power BI, and advanced statistical analysis, making him well-suited for data-driven roles.",
      "job_fit_analysis": {
        "reasoning": "**STRENGTHS:**\n- Strong technical skills match: Python, SQL, Tableau, Power BI all directly mentioned\n- Relevant educational background: Computer Science degree\n- Appropriate experience level: 5 years exceeds minimum requirement of 3 years\n- Advanced skills: Machine Learning mentioned as preferred skill\n\n**GAPS/WEAKNESSES:**\n- No explicit mention of Excel proficiency\n- Statistics knowledge could be more detailed\n\n**OVERALL RECOMMENDATION:** Yes\n- Excellent technical skill alignment (90% match)\n- Education and experience requirements met\n- Strong foundation for senior data analyst role\n- Fit percentage: 87%",
        "fit_score": 0.87,
        "recommendation": "Yes"
      },
      "analysis_timestamp": "2025-08-29T13:04:23.456789",
      "processing_time_seconds": 3.2
    }
  ]
}

📈 Detailed Analysis Report

====================================================
📄 ANALYSIS RESULT: john_smith_cv.pdf
====================================================

🎯 JOB FIT SUMMARY:
   Recommendation: Yes
   Fit Score: 87%
   Processing Time: 3.20s

📝 CANDIDATE SUMMARY:
   John Smith is an experienced data analyst with 5 years of professional 
   experience and a strong educational background from MIT. His technical 
   expertise includes Python, SQL, Tableau, Power BI, and advanced 
   statistical analysis, making him well-suited for data-driven roles.

🤔 AI REASONING:
   STRENGTHS:
   - Strong technical skills match: Python, SQL, Tableau, Power BI all directly mentioned
   - Relevant educational background: Computer Science degree
   - Appropriate experience level: 5 years exceeds minimum requirement of 3 years
   - Advanced skills: Machine Learning mentioned as preferred skill

   GAPS/WEAKNESSES:
   - No explicit mention of Excel proficiency
   - Statistics knowledge could be more detailed

   OVERALL RECOMMENDATION: Yes
   - Excellent technical skill alignment (90% match)
   - Education and experience requirements met
   - Strong foundation for senior data analyst role
   - Fit percentage: 87%

🔧 Technical Details

🤖 AI Model Specifications

Model: Qwen/Qwen2.5-1.5B-Instruct
Size: ~3GB download
Type: Instruction-tuned large language model
Optimization: FP16 precision for GPU, FP32 for CPU
Context Length: 8192 tokens

🏗️ Core Components

SimpleCVAnalyzer Class

class SimpleCVAnalyzer:
    def __init__(self):
        # Initialize Qwen2.5 model and tokenizer
        
    def extract_text_from_pdf(self, pdf_path: str) -> str:
        # PDF text extraction using PyPDF2
        
    def extract_cv_information(self, cv_text: str) -> Dict:
        # AI-powered structured information extraction
        
    def generate_candidate_summary(self, cv_info: str) -> str:
        # Professional summary generation
        
    def reason_job_fit(self, cv_info: str, candidate_summary: str) -> Dict:
        # Job fit analysis with detailed reasoning
        
    def analyze_cv(self, pdf_path: str) -> Dict:
        # Complete analysis pipeline
        
    def analyze_multiple_cvs(self, cv_folder: str) -> List[Dict]:
        # Batch processing with ranking

🛡️ Anti-Hallucination Measures

Strict Prompt Engineering

prompt = """
IMPORTANT: Only extract information that is EXPLICITLY mentioned in the CV text.
Do NOT add, assume, or infer any information that is not clearly stated.

CRITICAL RULES:
- If information is not in the CV, write "Not mentioned" or "Not provided"
- Do not assume or add skills that are not explicitly listed
- Stick strictly to what is written in the document
"""

Conservative Generation Parameters

temperature=0.3,          # Lower creativity, higher accuracy
top_p=0.6,               # More focused token selection
repetition_penalty=1.1    # Reduce repetitive responses

Fact-Based System Prompts

system_prompt = """
You are a professional HR assistant that analyzes CVs with strict accuracy. 
You MUST only extract and use information that is explicitly stated in the 
provided documents. Never assume, infer, or add information that is not 
clearly written.
"""

⚡ Performance Optimizations

GPU Detection: Automatic CUDA utilization when available
Memory Management: Efficient tensor handling and cleanup
Batch Processing: Optimized for multiple CV analysis
Error Recovery: Graceful handling of processing failures

💡 Best Practices

📄 CV Preparation

PDF Quality: Use text-based PDFs (not scanned images)
File Naming: Use descriptive filenames (e.g., "john_smith_cv.pdf")
Content Structure: Well-organized CVs produce better results
Language: English CVs work best with current model

🎯 Job Requirements Setup

Be Specific: Detailed requirements lead to better matching
Use Standard Terms: Common industry terminology works best
Balance Requirements: Include must-haves and nice-to-haves
Update Regularly: Keep job templates current with market needs

🔍 Analysis Interpretation

Fit Scores:
- 80-100%: Excellent match, strong candidate
- 65-79%: Good match, worth interviewing
- 40-64%: Possible match, review carefully
- Below 40%: Poor match, likely not suitable
Review AI Reasoning: Always read the detailed reasoning for context
Consider Soft Skills: Technical fit is important, but not everything
Use as Initial Screening: Supplement with human review for final decisions

🔧 System Optimization

Hardware Recommendations:
- Minimum: 4GB RAM, Intel i5 or equivalent
- Recommended: 8GB+ RAM, GPU with 4GB+ VRAM
- Optimal: 16GB+ RAM, RTX 3060 or better
Batch Size Guidelines:
- Small batches: 1-5 CVs for testing
- Medium batches: 10-20 CVs for regular use
- Large batches: 50+ CVs for bulk processing

🔨 Troubleshooting

❌ Common Issues

1. Model Download Failures

Error: Failed to download model

Solutions:

Check internet connection
Ensure sufficient disk space (3GB+)
Try running again (downloads resume automatically)
Use VPN if in restricted region

2. Memory Errors

RuntimeError: CUDA out of memory

Solutions:

Close other applications
Use CPU instead of GPU: Set device="cpu"
Reduce batch size in code
Restart Python session

3. PDF Reading Errors

Error reading PDF: PdfReadError

Solutions:

Ensure PDF is not password-protected
Try converting to different PDF format
Check file permissions
Use text-based PDFs (not scanned images)

4. Import Errors

ImportError: No module named 'transformers'

Solutions:

Activate virtual environment
Install requirements: pip install -r requirements.txt
Check Python version (3.8+ required)
Reinstall dependencies if needed

🔧 Performance Issues

Slow Processing

Check system resources (Task Manager/Activity Monitor)
Use GPU if available (automatic detection)
Process smaller batches
Upgrade hardware if regularly processing large volumes

Inconsistent Results

Check CV quality (text-based vs. scanned)
Verify job requirements are specific and clear
Update model if using cached version
Review AI reasoning for patterns

📞 Getting Help

Check logs for detailed error messages
Review requirements.txt for dependency versions
Test with sample CVs to isolate issues
Update dependencies if problems persist

🛠️ Development

🔧 Code Structure

Main Application (`main.py`)

Interactive job setup
CV source selection
Analysis execution
Result presentation

Core Analyzer (`analyzer/processors/simple_cv_analyzer.py`)

AI model initialization
PDF text extraction
Information extraction pipeline
Job fit analysis

Alternative Interface (`analyzer/cv_analyzer.py`)

Command-line interface
Batch processing utilities
Result formatting

🧪 Testing

# Test with sample CVs
python main.py
# Select option 3 (default folder)
# Use test CVs in cv/ folder

# Command line testing
cd analyzer
python cv_analyzer.py --file ../cv_test_1/test.pdf --detailed

🔄 Adding New Features

New Job Templates: Add to job_templates dictionary in main.py
Custom Scoring: Modify reason_job_fit method
Output Formats: Extend result saving functions
Model Updates: Change model_name in SimpleCVAnalyzer

📦 Dependencies Management

# Update requirements
pip freeze > requirements.txt

# Install development dependencies
pip install jupyter notebook matplotlib seaborn

# Create minimal requirements
pip install torch transformers PyPDF2 pathlib2

🚀 Future Enhancements

🔮 Planned Features

Multi-language Support: Support for non-English CVs
Resume Parsing APIs: RESTful API for system integration
Web Interface: Browser-based CV upload and analysis
Database Integration: Store and track candidate analyses
Advanced Filtering: Complex query-based candidate filtering
Custom Model Training: Fine-tune models for specific industries

🎯 Optimization Roadmap

Model Quantization: Reduce memory usage further
Async Processing: Concurrent CV analysis
Caching System: Store processed results for quick retrieval
Auto-scaling: Dynamic resource allocation based on workload

📄 License

This project is developed for educational and research purposes.

📚 Third-Party Components

Qwen2.5: Apache 2.0 License
Transformers: Apache 2.0 License
PyPDF2: BSD License
Other dependencies: Various open-source licenses

⚖️ Usage Terms

✅ Educational use
✅ Research purposes
✅ Internal business use
❌ Commercial redistribution without permission
❌ Using for discriminatory hiring practices

🙏 Acknowledgments

🏆 Technologies Used

Qwen2.5 - Advanced language model by Alibaba
Hugging Face Transformers - Model hosting and inference
PyPDF2 - PDF text extraction
PyTorch - Deep learning framework

👥 Contributors

Development Team - Core system implementation
AI Research Community - Model development and optimization
Open Source Community - Supporting libraries and tools

📞 Support & Contact

📊 Version: 1.0.0
📅 Last Updated: August 29, 2025
🔧 Compatibility: Python 3.8+ | Windows/macOS/Linux
💾 Model Size: 3GB | ⚡ Min RAM: 4GB | 🚀 Recommended RAM: 8GB+

Happy CV analyzing! 🎯✨

Name		Name	Last commit message	Last commit date
analyzer		analyzer
cv		cv
cv_test_1		cv_test_1
.gitattributes		.gitattributes
.gitignore		.gitignore
README.md		README.md
cv_analysis_batch_20250829_130423.json		cv_analysis_batch_20250829_130423.json
cv_analysis_results_20250829_130423.json		cv_analysis_results_20250829_130423.json
main.py		main.py
requirements.txt		requirements.txt

Latest commit

History

Folders and files

Repository files navigation

🎯 AI-Powered CV Analysis System

📋 Table of Contents

🎯 Overview

🎯 Target Use Cases

✨ Key Features

🤖 AI-Powered Analysis

📄 Document Processing

🎯 Job Matching System

🛡️ Anti-Hallucination Technology

📊 Comprehensive Reporting

🏗️ System Architecture

🔄 Processing Pipeline

🚀 Quick Start Guide

📋 Prerequisites

⚡ Installation

🎬 First Run

📖 Detailed Usage

🎯 Interactive Mode (Recommended)

🖥️ Command Line Mode

📝 Job Template Examples

Data Analyst Template

Software Engineer Template

⚙️ Configuration

🎛️ Job Requirements Configuration

Custom Job JSON Format

🔧 AI Model Configuration

💾 Memory Optimization

📊 Output Examples

📱 Console Output

📄 JSON Output Structure

📈 Detailed Analysis Report

🔧 Technical Details

🤖 AI Model Specifications

🏗️ Core Components

SimpleCVAnalyzer Class

🛡️ Anti-Hallucination Measures

⚡ Performance Optimizations

💡 Best Practices

📄 CV Preparation

🎯 Job Requirements Setup

🔍 Analysis Interpretation

🔧 System Optimization

🔨 Troubleshooting

❌ Common Issues

1. Model Download Failures

2. Memory Errors

3. PDF Reading Errors

4. Import Errors

🔧 Performance Issues

Slow Processing

Inconsistent Results

📞 Getting Help

🛠️ Development

🔧 Code Structure

Main Application (main.py)

Core Analyzer (analyzer/processors/simple_cv_analyzer.py)

Alternative Interface (analyzer/cv_analyzer.py)

🧪 Testing

🔄 Adding New Features

📦 Dependencies Management

🚀 Future Enhancements

🔮 Planned Features

🎯 Optimization Roadmap

📄 License

📚 Third-Party Components

⚖️ Usage Terms

🙏 Acknowledgments

🏆 Technologies Used

👥 Contributors

📞 Support & Contact

About

Resources

Stars

Watchers

Forks

Releases

Main Application (`main.py`)

Core Analyzer (`analyzer/processors/simple_cv_analyzer.py`)

Alternative Interface (`analyzer/cv_analyzer.py`)