🎯 Job Application Automation System

A self-hosted AI pipeline that automatically discovers, scores, and tailors job applications — built with n8n, Claude API, PostgreSQL, and Docker.

---

What It Does

Every 2 hours, the system automatically:

1. Discovers jobs from RSS feeds (Jobicy, HN Who's Hiring)
2. Extracts structured data from each job description using Claude AI
3. Scores each job 0–100 against your resume and profile
4. Filters to only jobs scoring ≥ 70 (configurable threshold)
5. Tailors ATS-optimized resume bullets for the top 2 jobs
6. Writes a personalized cover letter for the top 1 job
7. Stores everything in PostgreSQL
8. Sends a daily digest by email and Telegram at 8AM

The system never auto-submits applications — you review and apply manually.

---

Architecture

+----------------------------------------------------------+
|              LOCAL DOCKER ENVIRONMENT                    |
|                                                          |
|  +-------------+   +--------------+   +-------------+   |
|  |  job-n8n    |   | job-postgres |   |job-gotenberg|   |
|  | port: 5678  |<->|  port: 5432  |   | port: 3001  |   |
|  | Workflows   |   | Job storage  |   | PDF service |   |
|  +------+------+   +--------------+   +-------------+   |
|         |                                                |
+---------+------------------------------------------------+
          | External API calls
          v
+----------------------------------------------------------+
|  Claude API  |  RSS Feeds  |  Gmail SMTP  |  Telegram   |
+----------------------------------------------------------+

---

Tech Stack

Layer	Technology	Purpose
Orchestration	n8n (self-hosted)	Visual workflows, scheduling, node connections
AI / LLM	Claude Haiku 4.5	Job extraction, scoring, resume tailoring, cover letters
Database	PostgreSQL 15	Persistent storage for jobs, scores, materials
Containers	Docker + Compose	Isolated local services, one-command startup
Email	Gmail SMTP	Daily digest via App Password
Messaging	Telegram Bot	Real-time alerts and daily summaries
Job Sources	RSS Feeds	Jobicy, HN Who's Hiring (public feeds, no scraping)

---

Three Workflows

Workflow 01 — Job Discovery & Analysis (every 2 hours)

RSS Feeds → Compute Hashes (dedup) → Claude: Extract JD → Claude: Score Fit
→ Filter ≥70 → Top 2: Tailor Resume → Top 1: Cover Letter → PostgreSQL

Workflow 02 — Email Alert Processor (every 5 min)

Gmail IMAP → Extract URLs from alerts → Same analysis pipeline as WF01

Workflow 03 — Daily Digest (8AM daily)

Query PostgreSQL → Format HTML Email → Send via Gmail SMTP + Telegram Bot

---

Cost

Task	Items/run	Cost/call	Run total
Claude Extract JD	~55	$0.00004	$0.0022
Claude Match Score	~55	$0.00004	$0.0022
Claude Tailor Resume	top 2	$0.0002	$0.0004
Claude Cover Letter	top 1	$0.0003	$0.0003
Total per run			~$0.005

~$1.80/month for 12 runs/day. Infrastructure: $0 (local Docker).

---

Quick Start

1. Prerequisites

• Docker Desktop installed
• Anthropic API key (console.anthropic.com)
• Gmail App Password (myaccount.google.com → Security → App Passwords)
• Telegram Bot token (optional — @BotFather)

2. Setup

git clone https://github.com/YOUR_USERNAME/job-automation
cd job-automation

# Copy env template and fill in your values
cp .env.example .env
nano .env  # or open in any editor

# Add your resume
cp resumes/base-resume.txt resumes/my-resume.txt
# Edit with your actual details

# Start everything
chmod +x start.sh stop.sh db.sh update.sh
./start.sh

3. Open n8n

http://localhost:5678

Log in → Import the 3 workflow JSON files from workflows/ folder → Activate all 3.

4. Add Credentials in n8n

• PostgreSQL: host=postgres, db=jobautomation, user=jobuser
• SMTP: Gmail with App Password, port 587
• IMAP: Gmail with App Password, port 993
• Telegram: Bot token from @BotFather

---

Daily Usage

# Check new unreviewed jobs
./db.sh new

# Top jobs by score
./db.sh top

# Get cover letter for a job
./db.sh cover <job_id>

# Get tailored resume for a job
./db.sh resume <job_id>

# Mark job as applied
./db.sh apply <job_id>

# Open DB shell
./db.sh shell

# Update n8n to latest
./update.sh

---

Database Schema (key fields)

Column	Description
job_id	URL hash — prevents duplicates
match_score	Claude's 0–100 fit score
recommendation	apply / maybe-apply / skip
tailored_resume_text	ATS-optimized resume for this job
cover_letter	Personalized cover letter
missing_skills	Skills gap identified by Claude
strong_matches	Your strengths for this role
visa_note	remote-friendly / EU-based / requires-sponsorship

---

Human-in-the-Loop Design

This system never auto-submits applications. It only:

• ✅ Discovers relevant jobs
• ✅ Scores fit against your profile
• ✅ Prepares tailored materials
• ✅ Notifies you daily

You decide which jobs to apply to and click Apply yourself.

---

Project Structure

job-automation/
├── docker-compose.yml       # All services defined here
├── .env.example             # Template (copy to .env)
├── init.sql                 # PostgreSQL schema
├── start.sh                 # Start all containers
├── stop.sh                  # Stop all containers
├── update.sh                # Update n8n to latest
├── db.sh                    # Database management CLI
├── resumes/
│   └── base-resume.txt      # Your resume template
└── workflows/
    ├── 01-job-discovery.json
    ├── 02-email-alert-processor.json
    └── 03-daily-digest.json

---

Built By

Deepak Kushwaha — Data Scientist & ML Engineer
MSc Data Science, University of Naples Federico II
🏆 1st Place, Google Challenge Campania 2026 (STEM Category)

---

License

MIT — feel free to adapt for your own job search.