The latest articles on DEV Community by Ihor Bezruchko (@ihorbbt). https://dev.to/ihorbbt https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3978194%2F1f54f8fd-05ea-42cc-ac5c-4709b499dbd3.png https://dev.to/ihorbbt en Ihor Bezruchko Wed, 10 Jun 2026 18:21:33 +0000 https://dev.to/ihorbbt/i-built-a-freelance-job-hunting-automation-on-n8n-heres-everything-i-learned-4mi https://dev.to/ihorbbt/i-built-a-freelance-job-hunting-automation-on-n8n-heres-everything-i-learned-4mi <p>I'm a 17-year-old IT student from Luxembourg. A few months ago I got tired of spending 2-3 hours a day manually browsing Upwork, Malt, and Freelancer looking for projects.</p> <p>So I built an automation system that does it for me — 24/7, on a Raspberry Pi 3.</p> <p>Here's what it does, how I built it, and every painful lesson I learned along the way.</p> <h2> What the system does </h2> <ul> <li>Scans Upwork, Malt, and Freelancer every 30 minutes</li> <li>Scores each job 0–100 with AI based on my profile</li> <li>Generates proposals in English, French, and German</li> <li>Sends the best jobs to Telegram with inline A/B buttons</li> <li>Tracks which proposal style gets more replies</li> <li>Sends daily stats and weekly market trend reports</li> <li>Reminds me to follow up after 3 days</li> </ul> <h2> The stack </h2> <ul> <li> <strong>n8n</strong> — self-hosted workflow automation (Docker on Raspberry Pi 3)</li> <li> <strong>Groq API</strong> (Llama 3.1-8b-instant) — AI scoring and proposal generation</li> <li> <strong>Supabase</strong> — PostgreSQL database for jobs, proposals, clients</li> <li> <strong>SerpAPI</strong> — searching job boards via Google</li> <li> <strong>Apify</strong> — scraping Upwork listings</li> <li> <strong>Telegram Bot API</strong> — alerts and bot commands</li> <li> <strong>Cloudflare Tunnel</strong> — HTTPS for webhooks</li> </ul> <p>Total running cost: ~$5/month.</p> <h2> 7 workflows </h2> <p><strong>01 - Job Discovery</strong> — runs every 30 minutes, searches 10+ sources, deduplicates via Supabase unique constraint on URL</p> <p><strong>02 - Proposal Generator</strong> — AI scores the job, generates two proposal variants (formal vs hook-first), sends to Telegram with A/B buttons</p> <p><strong>03 - Follow-up Reminders</strong> — checks Supabase every 3 days for unanswered proposals</p> <p><strong>04 - CRM via Telegram</strong> — full client management through bot commands (/jobs, /stats, /clients)</p> <p><strong>05 - Market Intelligence</strong> — daily report: how many jobs found, average score, top platforms</p> <p><strong>06 - Trend Analysis</strong> — weekly report on what skills are trending in automation</p> <p><strong>07 - Lead Generation</strong> — finds companies actively using Zapier or Make who might want to switch to n8n</p> <h2> Lessons learned (the hard way) </h2> <p><strong>1. Cyrillic text breaks JSON body nodes silently</strong><br> If you have Cyrillic characters in a JSON body field with newlines, n8n throws a "Bad control character" error. Keep everything on a single line or use <code>\n</code> instead of real line breaks.</p> <p><strong>2. Multiple inputs into a Code node need a Merge node first</strong><br> When two HTTP Request nodes feed into one Code node, use a Merge node (Combine by Position) in between. Otherwise <code>$('NodeName')</code> throws "hasn't been executed" errors.</p> <p><strong>3. Cloudflare tunnel URL changes on every restart</strong><br> This breaks all Telegram webhooks. I solved it with a systemd service that restarts cloudflared and updates the <code>WEBHOOK_URL</code> environment variable in docker-compose automatically.</p> <p><strong>4. Deduplication belongs in the database, not in n8n</strong><br> Adding a UNIQUE constraint on <code>jobs.url</code> in Supabase and setting the insert node to "Continue on Fail" is cleaner than trying to check duplicates inside the workflow.</p> <p><strong>5. Code nodes must always return items</strong><br> If a Code node returns nothing (empty array or undefined), the workflow silently passes dummy data downstream. Always return <code>[]</code> or a proper item.</p> <p><strong>6. LLMs ignore "return only JSON" instructions</strong><br> Groq kept wrapping JSON in <br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> json markdown blocks. Fix: strip the backticks before parsing, or use separator-based output instead of JSON. --- ## A/B testing proposals This was the most interesting part to build. Instead of one proposal, the system now generates two variants with different tones: - **Variant A** — formal, structured, under 100 words - **Variant B** — hook-first, starts with the client's problem, under 70 words Telegram sends both with buttons. When I tap which one I sent, the result gets stored in Supabase. Over time I'll have data on which style works better for which type of job. --- ## GitHub All 7 workflow files (sanitized — API keys replaced with placeholders) are on GitHub: 👉 https://github.com/ihorbezruchko/n8n-freelance-autopilot If you want the full working setup with the complete configuration guide: 👉 https://n8n-freelance-autopilot.netlify.app --- Happy to answer questions about any part of the implementation. What would you build differently? </code></pre> </div> ai productivity javascript career