Skip to content

🎯 START HERE - Episode 02 Quick Start Guide

Updated: October 1, 2025 Claude Sonnet 4.5 Released: September 29, 2025

🚀 What You're About to Build

An AI Resume Agent powered by Claude Sonnet 4.5 that:

✅ Transforms any resume into a perfect job match in 2 minutes ✅ Works for any profession (software, marketing, real estate, etc.) ✅ Achieves 3x higher response rates (12% → 44%) ✅ Gets you 6x more interviews (3 → 18 per 50 applications) ✅ Saves 72+ hours per job search

⏱️ Time Investment

  • Setup: 30 minutes (one-time)
  • Create experience DB: 1-2 hours (one-time, reusable forever)
  • Per resume: 2 minutes (ongoing)

ROI: After 5 resumes, you've saved more time than you invested.

📂 File Structure (What to Read & When)

02-resume-fine-tuner/
│
├── 🎯 START_HERE.md                    ← You are here
├── 📹 THE_FULL_SCRIPT.md               ← Read next (complete video script)
│
├── prompts/
│   ├── resume-fine-tuner-agent.md      ← Step 3: Copy this
│   └── ALL_CLAUDE_PROMPTS.md           ← Reference (all prompts)
│
└── templates/
    ├── experience-software-engineer.md  ← Step 4: Customize one of these
    ├── experience-marketing-manager.md
    └── experience-real-estate-agent.md

🎬 If You're Creating the Video

Read this order: 1. ✅ This file (START_HERE) - 5 min 2. 📹 THE_FULL_SCRIPT.md - 30 min (your production bible) 3. Choose profession template from templates/ - 10 min 4. Test the system before recording - 30 min 5. Start recording following the script

Total prep time: ~75 minutes before you hit record

🏃 If You Just Want to Use It (Not Record Video)

Step 1: Prerequisites (5 minutes)

Check if you have Claude Code:

claude --version

If not installed:

  • macOS/Linux: See Claude Code docs
  • Or watch Episode 01 of this series
  • Or comment on the YouTube video for help

Step 2: Clone Repository (2 minutes)

# Clone the repo
git clone https://github.com/anton-abyzov/ai-power.git

# Navigate to this episode
cd ai-power/episodes/02-resume-fine-tuner

# Verify files
ls -la

Step 3: Set Up the Agent (5 minutes)

Option A: Copy the agent file

# Create agents directory if it doesn't exist
mkdir -p ~/.claude/agents

# Copy agent
cp prompts/resume-fine-tuner-agent.md ~/.claude/agents/

Option B: Create manually 1. Open prompts/resume-fine-tuner-agent.md 2. Copy the entire content 3. Create a new file in your project: resume-fine-tuner-agent.md 4. Paste and save

Test it works:

claude
# Inside Claude:
@resume-fine-tuner help

Should show agent description and capabilities.

Step 4: Create Your Experience Database (1-2 hours, one-time!)

This is the most important step. Take your time.

Choose your template:

  • Software/Tech: templates/experience-software-engineer.md
  • Marketing/Growth: templates/experience-marketing-manager.md
  • Sales/Real Estate: templates/experience-real-estate-agent.md

Copy and customize:

# Example for software engineer
cp templates/experience-software-engineer.md experience-db/my-experience.md

# Open and edit with YOUR info
code experience-db/my-experience.md
# or
vim experience-db/my-experience.md
# or
nano experience-db/my-experience.md

What to include: ✅ Every role (current → past) ✅ Technologies, tools, frameworks ✅ Quantified achievements (with realistic metrics: 20-40%, not "10x") ✅ Team size, scale, impact ✅ Education, certifications ✅ Skills (categorized: expert → intermediate)

Pro tips:

  • Be honest (you'll discuss this in interviews)
  • Use realistic numbers (25% improvement > "revolutionary")
  • Include context (team size, scale, budget)
  • Organize chronologically (newest first)

Step 5: Find a Job & Generate Resume (2 minutes)

Find a real job: 1. Go to LinkedIn Jobs, Indeed, or company career page 2. Find a role that matches your experience 3. Copy the full job description 4. Save it: job-descriptions/[company-name]-[role].md

Generate resume:

# Start Claude
claude

# Inside Claude, use this prompt:
@resume-fine-tuner

Create a tailored resume for [JOB TITLE] at [COMPANY NAME]:

Read job description: job-descriptions/[company-name]-[role].md

Use experience database: experience-db/my-experience.md

Output file: outputs/[company-name]-[role]-2025-10-01.md

Emphasize:

- [Key skill 1 from job posting]
- [Key skill 2 from job posting]
- [Key skill 3 from job posting]
- [Key achievement type they want]

Ensure 90%+ keyword match while maintaining authenticity.

Wait 60-90 seconds while Claude:

  • Reads job description
  • Extracts keywords
  • Searches your experience DB
  • Matches relevant achievements
  • Writes tailored resume

Done! Resume saved to outputs/ folder.

Step 6: Convert to Desired Format (1 minute)

Claude outputs Markdown by default. Convert to:

Word (.docx) - Recommended:

pandoc outputs/your-resume.md -o outputs/your-resume.docx

PDF (from Word): 1. Open the .docx file in Word 2. File → Save As → PDF 3. Done!

Or use online converters (no installation):

  • Dillinger.io - Paste Markdown → Export as PDF/DOCX
  • StackEdit.io - Paste Markdown → Export
  • CloudConvert.com - Upload .md → Convert to PDF/DOCX

LaTeX (for academic resumes):

pandoc outputs/your-resume.md -o outputs/your-resume.tex

Then compile in Overleaf or with LaTeX installed locally.

🎯 Your First 30 Minutes

Realistic timeline:

00:00 - Install Claude Code (if needed)
00:05 - Clone repository
00:07 - Copy agent file
00:10 - Test agent works
00:12 - Choose experience template
00:15 - Start customizing experience DB
01:45 - Finish experience DB
01:47 - Find a real job posting
01:50 - Generate first resume
01:52 - Convert to PDF
01:55 - Review and iterate
02:00 - Apply to job!

First resume in 2 hours (including 1.5h experience DB setup). Every resume after that: 2 minutes.

💡 Pro Tips

Tip 1: Build Your Experience DB Thoroughly (Once)

  • Spend 1-2 hours making it comprehensive
  • Include EVERY achievement, technology, metric
  • The richer your DB, the better every resume will be
  • This is a one-time investment with infinite returns

Tip 2: Save Job Descriptions That Worked

# When you get an interview, save the JD
mkdir -p job-descriptions/interviews
cp [job-description].txt job-descriptions/interviews/
  • Pattern recognition: What keywords led to interviews?
  • Reuse for similar roles

Tip 3: Track Your Results

Create a simple CSV:

Date,Company,Role,Response,Interview,Offer,Notes
2025-10-01,TechCorp,Senior SWE,Yes,Yes,Pending,Great culture fit
2025-10-02,StartupXYZ,Staff Eng,No,-,-,Too senior

Helps you optimize over time.

Tip 4: Customize the Prompt for Each Job

Don't just paste job description. Add:

Emphasize:

- [Specific tech from JD]
- [Type of experience they want]
- [Company culture fit]

De-emphasize:

- [Irrelevant experience]

Tone: [Startup casual / Enterprise formal / Academic rigorous]

More specific = better results.

Tip 5: Use Variations for Different Seniority

Same job type, different levels:

  • Junior: Emphasize learning, growth, foundational skills
  • Mid: Emphasize delivery, technical depth, independence
  • Senior: Emphasize leadership, architecture, business impact
  • Staff+: Emphasize strategy, org-level impact, technical direction

Same experience DB, different emphasis.

🚨 Common Mistakes (Avoid These!)

Mistake 1: Rushing the Experience DB

❌ "I'll add more later" ✅ Take 1-2 hours to make it comprehensive NOW

Why: Every resume pulls from this. Garbage in = garbage out.

Mistake 2: Copying Job Description Blindly

❌ Just pasting JD without reading ✅ Analyze JD, identify key requirements, tell Claude what to emphasize

Why: Claude is smart, but your context helps.

Mistake 3: Not Tracking Results

❌ "I'll remember what worked" ✅ Track every application, response, interview

Why: Data tells you what to optimize.

Mistake 4: Using Unrealistic Metrics

❌ "Improved performance 10x" ✅ "Reduced latency 28% through caching"

Why: You'll have to defend metrics in interviews. Be honest.

Mistake 5: One Generic Experience DB

❌ Cramming everything into one file ✅ Organize: experience-db/software-eng.md, experience-db/devops.md

Why: Different career paths need different emphasis.

🎬 Video Production Path (If Recording)

Recording Order (Most Efficient):

Session 1: On-Camera (30 min)

  • Hook (0:00-0:45)
  • Closing (16:00-16:45)
  • Done with camera work!

Session 2: Browser Recordings (1 hour)

  • Model comparison pages
  • Benchmark pages
  • LinkedIn job searches
  • GitHub repo tour

Session 3: Terminal Recordings (30 min)

  • Setup commands
  • Agent creation
  • Format conversions

Session 4: Claude Demos (2 hours) - MOST IMPORTANT

  • Software engineer demo (uncut, real-time)
  • Marketing manager demo (uncut)
  • Real estate agent demo (uncut)
  • Show generated resumes

Session 5: Voiceover (1.5 hours)

  • Record following script
  • High energy, clear pronunciation
  • Record in sections

Session 6: Editing (6-8 hours)

  • Sync voiceover
  • Add text overlays
  • Animate metrics
  • Color grade
  • Export

Total: 12-15 hours for a professional 15-18 min video

📊 Success Criteria

You'll know it's working when:

✅ First resume generated in under 5 minutes ✅ Resume contains 90%+ of job keywords ✅ All achievements are from YOUR real experience (authentic) ✅ Response rate improves (track first 10 applications) ✅ You're spending 2 minutes per resume, not 2 hours

Iterate if:

  • Response rate is still low (<30%) → Add more keywords to experience DB
  • Resumes feel generic → Add more specific project details
  • ATS rejecting → Simplify formatting, use standard headers
  • Too long (>2 pages) → Prioritize recent experience, remove older roles

Essential:

Templates:

Examples:

❓ FAQ

Q: Does this work with ChatGPT or Gemini instead of Claude? A: Yes! The workflow works with any model. But Claude Sonnet 4.5 has:

  • 77.2% SWE-bench (vs 72.8% GPT-4)
  • Lower hallucination rate
  • Better instruction following
  • Same price as old Sonnet 4

Q: How long does the experience database take to create? A: 1-2 hours for a thorough job. But it's reusable forever and improves every resume.

Q: Can I use this for cover letters too? A: Absolutely! Same agent, different prompt:

@resume-fine-tuner

Create cover letter for [role] at [company]:
[JD]
Use: experience-db/my-experience.md
Length: 250-300 words
Tone: Professional but personable

Q: What if I don't have Claude Code access yet? A: You can: 1. Use Claude.ai web interface (copy prompts manually) 2. Wait for Claude Code access 3. Adapt to ChatGPT/Gemini (slight prompt modifications needed)

Q: Is this ATS-friendly? A: Yes! Markdown → PDF conversion creates clean, parseable resumes. Test with Jobscan.co to verify.

Q: How do I update my experience DB over time? A: Just edit the file! Add new roles, achievements, skills. Version control with git to track changes.

🚀 Next Steps

Right Now (Next 5 Minutes): 1. Verify Claude Code is installed: claude --version 2. Clone this repo: git clone https://github.com/anton-abyzov/ai-power.git 3. Read the full script: 📹 THE_FULL_SCRIPT.md

Today (Next 2 Hours): 1. Copy agent configuration 2. Choose and customize experience template 3. Generate your first tailored resume 4. Apply to a real job

This Week: 1. Apply to 10+ jobs using tailored resumes 2. Track results (responses, interviews) 3. Iterate based on data

Share Your Success:

  • Got an interview? Comment on YouTube!
  • Built an improvement? Open a PR!
  • Found it helpful? Star the repo ⭐

💪 You're Ready!

Everything you need is here:

  • The world's best AI model (Claude Sonnet 4.5)
  • Complete agent system
  • Templates for multiple professions
  • Full video production script
  • Step-by-step guide

The only missing ingredient? You taking action.

Remember:

"You won't be replaced by AI. You'll be replaced by someone who knows how to use AI."

You now know how.

Let's go build something incredible! 🚀

Questions?

  • 📧 Open a GitHub Issue
  • 💬 Comment on the YouTube video
  • 🤝 Join the community

Start here: 📹 THE_FULL_SCRIPT.md →