[SERVICES] — N8N & AI AUTOMATION
I build and harden n8n automations for production — and I have the receipts: Chatflowgate, a live multi-tenant gateway for n8n chat, and n8n running the agent workflows behind GrowthOS. Metro Manila, Philippines, working with teams worldwide.
[01] WORKFLOWS
Workflows that survive contact with real traffic — error handling, retries, idempotency and monitoring, not just nodes that work in the editor. I run n8n in production behind GrowthOS's community-discovery agents.
[02] SECURITY
n8n's chat trigger exposes a bare public webhook — fine for demos, a liability in production. I built Chatflowgate, a multi-tenant gateway that adds tenant auth, origin checks, rate limiting and conversation history in front of it.
[ READ THE ARCHITECTURE → ][03] AI AGENTS
Wiring LLMs into n8n the way I wire them into products at my day job — copilots, RAG retrieval and multi-agent flows for a national-scale fintech. Guardrails and evals included, because automation that hallucinates is a liability.
[ PRODUCTION RAG NOTES → ][04] RESCUE
A workflow someone built that nobody understands, breaking quietly at 3am. I audit it, document it, harden it — or rebuild it small enough that your team can own it.
n8n's editor makes the first 80% look easy — the paid-for part is the last 20%: error handling, security around public webhooks, rate limits, idempotent retries, and knowing when a workflow should be code instead. I hand over documented workflows your team can maintain, not a black box.
Yes — that exact problem is why I built Chatflowgate. The chat trigger's public webhook gets a gateway in front of it: per-tenant site keys, origin allowlists, rate limiting and stored conversation history, with the n8n webhook URL kept server-side.
Yes. My day job is LLM copilots and RAG pipelines at a national-scale fintech; I bring the same discipline — evaluation sets, guardrails, fallbacks for when the model is wrong — to n8n automations so they are trustworthy enough to run unattended.
Contract and remote from GMT+8 (full APAC overlap, mornings with Europe, evenings with the US). Scoped builds, audits, or ongoing support. Start with the contact form or email — I usually reply within 24 hours.