AI Cold Outreach Pipeline

Automated pipeline that scrapes AI/ML hiring and funding posts from X.com, LinkedIn, and news sites, finds relevant contacts, drafts personalized cold emails, and sends them -- all on autopilot.

Quick Start

1. Install dependencies:

   pip install -r requirements.txt
   playwright install chromium

2. Configure credentials:

   cp .env.example .env
   # Edit .env with your actual credentials

3. Set up Notion:

   • Create an integration at https://www.notion.so/my-integrations
   • Create a page and connect the integration to it
   • The pipeline will auto-create the databases on first run

4. First run (manual):

   python main.py --run-now

   • Browser windows will open for LinkedIn and X.com login on first run
   • Log in manually; sessions are saved for future runs

Headless / no browser popup mode

The pipeline now runs headless by default (scraping.headless: true), so no browser window opens during normal runs.

For authenticated X/LinkedIn scraping in headless mode, seed sessions once:

python setup_sessions.py --platform both

After that, all runs stay headless:

python demo.py
python main.py --run-now

Find email from a LinkedIn profile URL

If LinkedIn People search is blocked/unstable for your account, use direct profile lookup:

python find_email_from_linkedin_profile.py --linkedin-url "https://www.linkedin.com/in/USERNAME/"

Recommended for best accuracy:

python find_email_from_linkedin_profile.py \
  --linkedin-url "https://www.linkedin.com/in/USERNAME/" \
  --company "OpenAI" \
  --domain "openai.com"

If LinkedIn shows security verification, run with visible browser:

python find_email_from_linkedin_profile.py --linkedin-url "..." --headful

5. Install as daily service (runs at 6 AM):

   bash install_service.sh

Architecture

• Scrapers: Playwright (X.com, LinkedIn), BeautifulSoup (TechCrunch, Google News)
• Processing: Rule-based classifier + regex extractor + two-layer dedup
• Research: LinkedIn people search + email pattern guessing + SMTP verification
• Outreach: Groq LLM drafting + Gmail SMTP sending
• Storage: Notion databases (Leads, Contacts, Outreach)
• Orchestration: CrewAI agents + macOS launchd scheduling

Run frontend locally

The marketing site (ColdApply landing page) runs separately from the Python pipeline.

Why you see Can't resolve 'tailwindcss': The bundler resolves modules from the current working directory. If you run npm run dev from the project root (custom-cold-applying/) instead of from frontend/, it looks for tailwindcss in the root and fails. Fix: run from frontend/ or use the root scripts below.

Option A — From project root (recommended if your IDE opens the whole repo)

From custom-cold-applying/:

1. Install frontend dependencies (first time only):

   npm run frontend:install

2. Start the dev server:

   npm run frontend

3. Open http://localhost:3000 in your browser.

Option B — From the frontend folder

cd frontend
npm install          # first time only
npm run dev

Then open http://localhost:3000.

Other commands (from inside frontend/): npm run build, npm run start

Cost

$0. Everything uses free tiers or open-source tools.