I build and rescue n8n automations, and I run my own production instance rather than only building for other people: a self-hosted n8n server on a Hetzner VPS, 103 workflows, 74 of them live 24/7, plus around 20 Telegram bots, Docker on Linux.
What I take on
- New builds. Webhook and API integrations, LLM/agent steps with validated structured output, CRM and spreadsheet sync, scheduled scrapers, data pipelines.
-
- Rescues. A workflow that runs green but produces wrong results. This is most of what I fix, and the failures are almost always silent: a provider daily token cap that stamps a default value on every row while reporting success; a paginated API that quietly returns the first 250 of 1001 rows; an IF node checking the wrapper instead of the payload; a batch prompt where the model shifts its indices and every item gets its neighbour’s answer.
-
- Hardening. Explicit failure paths, a second provider behind the first, and a check that measures the output rather than the exit code.
- Also: Telegram bots including voice, Next.js/React front ends, Flutter mobile apps.
- How I work
Fixed price agreed before I start. I ask questions before I build - most failed projects fail in the brief, not in the code. If something in your stack is outside what I have shipped, I say so up front rather than discovering it on your project.
Based in Spain, remote, working in English.
My code and portfolio are on GitHub under the name Alexolimb - I cannot post links yet as a new account here, but I will send them on request, or you can find them from my profile.
Happy to walk through one of my live workflows on a call so you can judge the build quality before committing to anything.