PersonaGenerator

Generate comprehensive, AI-powered audience personas from LinkedIn profiles.

Overview

PersonaGenerator analyzes LinkedIn profiles and uses AI to synthesize detailed composite personas that include:

• Demographic & Professional Profile - Roles, industries, career stage, education
• Goals & Motivations - What drives them professionally and personally
• Pain Points & Challenges - Problems they face and constraints they work under
• Behaviors & Habits - Daily routines, decision-making style, technology adoption
• Communication Preferences - Tone, formats, detail level, reading time
• Content Engagement Patterns - Topics of interest, sharing behavior, engagement triggers
• Skills & Expertise - Core competencies and learning priorities
• Content Testing Framework - Actionable scores to test content relevance

Perfect For

• Content Creators: Test if your content will resonate with your target audience
• Marketers: Understand your ideal customer profile at a deep level
• Product Managers: Validate feature ideas against user personas
• Sales Teams: Better understand decision-makers and their pain points
• Founders: Define and understand your target market

Features

✅ Scrapes public LinkedIn profile data (no API key required) ✅ Uses AI to synthesize composite personas from multiple profiles ✅ Generates actionable markdown personas with 12 detailed sections ✅ Supports multiple AI models (Qwen/free, GPT-4, Claude, Gemini) ✅ Saves raw profile data for future reference ✅ CLI-friendly for automation

Installation

Prerequisites

1. Python 3.8+
2. Chrome browser (for Selenium)
3. ChromeDriver - Install with:

   brew install chromedriver

4. UnifiedLLMClient - Must be available at ~/Development/Scripts/UnifiedLLMClient

Setup

# Navigate to directory
cd ~/Development/Scripts/PersonaGenerator

# Install dependencies
pip3 install -r requirements.txt

# Copy environment template
cp .env.example .env

# Add your API keys to .env
# For free option, use Qwen (no API key needed)

Usage

Basic Usage

# From a file with URLs (one per line)
python3 persona_generator.py --urls sample_urls.txt

# From command line (comma-separated)
python3 persona_generator.py --urls "https://linkedin.com/in/user1,https://linkedin.com/in/user2,https://linkedin.com/in/user3"

Advanced Options

# Custom output directory
python3 persona_generator.py --urls urls.txt --output ./personas

# Custom persona name
python3 persona_generator.py --urls urls.txt --name "enterprise_cto_persona"

# Full example
python3 persona_generator.py \
  --urls executive_urls.txt \
  --output ./output/executives \
  --name c_suite_persona

Input Format

Create a text file with LinkedIn URLs (one per line):

# executive_urls.txt
https://www.linkedin.com/in/sarah-chen-tech-vp/
https://www.linkedin.com/in/john-smith-cto/
https://www.linkedin.com/in/maria-garcia-engineering-director/
https://www.linkedin.com/in/david-kumar-head-of-product/
https://www.linkedin.com/in/jennifer-liu-vp-operations/

Tips:

• Include 3-10 profiles for best results
• Choose profiles that represent your target audience
• More profiles = more accurate composite persona
• Profiles should share similar characteristics (role, industry, seniority)

Output

PersonaGenerator creates two files:

1. {name}_profiles.json

Raw scraped data from all LinkedIn profiles for reference.

2. {name}.md

Complete persona document with 12 sections:

1. Persona Overview - Name, archetype, summary
2. Demographics & Professional Profile - Roles, industries, education
3. Goals & Motivations - What they're trying to achieve
4. Pain Points & Challenges - Problems they face
5. Behaviors & Habits - Daily routines, decision-making
6. Communication Preferences - Tone, formats, reading time
7. Content Engagement Patterns - Topics, sharing behavior
8. Professional Context - Reporting structure, buying authority
9. Skills & Expertise - Competencies and knowledge areas
10. Content Testing Framework - Relevance, engagement, action scores
11. Example Content That Resonates - Specific headlines/topics
12. Anti-Patterns - What NOT to do

Example Output

# Persona: Tech Executive

## 1. PERSONA OVERVIEW

**Name**: Tech Executive
**Archetype**: Strategic Technical Leader
**One-line Summary**: Senior technology leaders who balance technical excellence with business strategy, leading engineering teams at high-growth companies.

## 2. DEMOGRAPHICS & PROFESSIONAL PROFILE

**Typical Roles/Titles**:
- VP of Engineering
- Chief Technology Officer (CTO)
- Head of Product Engineering
- Director of Engineering
- VP of Technology

**Industries**: SaaS, fintech, e-commerce, enterprise software...

[... continues with 10 more sections ...]

How It Works

┌─────────────────────┐
│  LinkedIn URLs      │
│  (input file)       │
└──────────┬──────────┘
           │
           ▼
┌─────────────────────┐
│  LinkedIn Scraper   │
│  (Selenium)         │
│  - Name             │
│  - Headline         │
│  - About            │
│  - Experience       │
│  - Education        │
│  - Skills           │
│  - Recent posts     │
└──────────┬──────────┘
           │
           ▼
┌─────────────────────┐
│  Profile Data       │
│  (JSON)             │
└──────────┬──────────┘
           │
           ▼
┌─────────────────────┐
│  AI Synthesizer     │
│  (UnifiedLLMClient) │
│  - Pattern analysis │
│  - Persona creation │
└──────────┬──────────┘
           │
           ▼
┌─────────────────────┐
│  persona.md         │
│  (Output)           │
└─────────────────────┘

AI Model Options

Configure in .env or via UnifiedLLMClient:

• qwen (default, free) - Good quality, no API key needed
• gpt-4o - High quality, requires OpenAI API key
• claude-3-5-sonnet - Excellent quality, requires Anthropic API key
• gemini-1.5-pro - Good quality, requires Google API key

Best Practices

For Better Personas:

1. Use 5-10 profiles - More data = better synthesis
2. Choose similar profiles - Same industry, role level, or target segment
3. Review the output - AI-generated, but should be human-validated
4. Update regularly - Personas should evolve with your market

For Content Testing:

Use the generated persona to evaluate content:

1. Relevance Check: Does this solve their problems?
2. Engagement Check: Would they read, like, or share?
3. Action Check: Would they take the next step?
4. Red Flag Check: Does it trigger any anti-patterns?

Limitations

• Public Data Only: Can only scrape publicly visible LinkedIn information
• LinkedIn Changes: Scraper may break if LinkedIn changes their HTML structure
• Rate Limiting: Be respectful of LinkedIn - don't scrape too many profiles at once
• Authentication: Does not use LinkedIn API (which requires auth)
• Accuracy: AI synthesis is based on patterns, not guaranteed predictions

Troubleshooting

ChromeDriver Not Found

brew install chromedriver
# Or download from: https://chromedriver.chromium.org/

UnifiedLLMClient Not Found

# Ensure UnifiedLLMClient is installed at:
~/Development/Scripts/UnifiedLLMClient

LinkedIn Blocking

If you get blocked:

• Wait a few hours before retrying
• Use fewer profiles per session
• Consider using LinkedIn Premium
• Use VPN if needed

Selenium Errors

# Update Selenium
pip3 install --upgrade selenium webdriver-manager

# Try non-headless mode for debugging
# Edit linkedin_scraper.py: headless=False

Examples

Example 1: SaaS Executive Persona

python3 persona_generator.py \
  --urls saas_execs.txt \
  --name saas_executive_persona

Example 2: Developer Persona

python3 persona_generator.py \
  --urls developer_profiles.txt \
  --name full_stack_developer_persona

Example 3: Marketing Leader Persona

python3 persona_generator.py \
  --urls marketing_leaders.txt \
  --output ./marketing_personas \
  --name cmo_persona

Integration with Other Tools

Use with AgenticContentGenerator

Generated personas can be used to test content from AgenticContentGenerator:

# 1. Generate persona
cd ~/Development/Scripts/PersonaGenerator
python3 persona_generator.py --urls exec_urls.txt --name exec_persona

# 2. Use persona to evaluate content
cd ~/Development/Scripts/AgenticContentGenerator
# Reference persona.md when reviewing generated topics/documents

Use with ContentScorer

Add persona-based scoring criteria to evaluate content fit.

Project Structure

PersonaGenerator/
├── README.md                    # This file
├── persona_generator.py         # Main CLI script
├── linkedin_scraper.py          # Selenium-based LinkedIn scraper
├── persona_synthesizer.py       # AI-powered persona synthesis
├── requirements.txt             # Python dependencies
├── .env.example                 # Environment template
├── .gitignore                   # Git ignore rules
├── sample_urls.txt              # Example input file
└── output/                      # Generated personas (auto-created)
    ├── {name}_profiles.json     # Raw profile data
    └── {name}.md                # Persona document

Contributing

Suggestions and improvements welcome! This tool is part of the Qwilo content generation ecosystem.

License

MIT License - Use freely for personal and commercial projects.

Related Tools

Part of the Qwilo content generation toolkit:

• UnifiedLLMClient - AI provider abstraction
• AgenticContentGenerator - Multi-agent content pipeline
• WritingStylePromptGenerator - Personal writing style capture

---

Created by: Stephen Sklarew Part of: Qwilo Content Generation Ecosystem Repository: ~/Development/Scripts/PersonaGenerator