Jor Ferraro

n8n Automation Engineer — Production Systems, Not Demos

jor@cdx.la n8n.cdx.la

I build production-grade automation systems that solve real business problems. 13+ years in broadcast post-production, now focused on n8n orchestration, LLM intelligence, and self-hosted infrastructure.

Three systems running in production: daily cultural automation (6 months, atmospheric data + Figma API), offer framework tool (Hormozi methodology via LLM), and streaming overlay generator (real-time AI descriptions). Plus one complex 27-node pipeline built as reference architecture, currently deployed as manual MVP.

n8n PostgreSQL Docker / Traefik OpenRouter Anthropic Claude Python Cloudflare Pages Figma API Notion API
01 Live — 90 days

Diario Sensorial

Automated daily cultural artifact for Buenos Aires

The Problem

Content creation is manual, inconsistent, and doesn't scale. Cities generate endless cultural data — weather, traffic, time of day — but no one's building systems that listen and respond automatically.

The Solution

Fully automated daily publication system that reads Buenos Aires atmospheric data (weather API + Google Directions), maps it to a curated musical matrix, generates context-aware copy via LLM, and outputs 5 Instagram slides daily at 10AM. Zero manual intervention after approval.

6 mo
Running Since Nov 2024
14
n8n Nodes
Daily
10AM Automated
Automation
n8n + cron
LLM
Sonnet 4
Data Sources
Weather + Directions APIs
Content DB
Notion (tracks + history)
Design
Figma API rendering
Distribution
Instagram Stories

Workflow

Cron 10AM Weather API Traffic API Build Mood Notion: Get Tracks OpenAI Notion: Create Draft Human Approval Figma API temp/humidity traffic mood track 5 texts approved 5 slides

The Mood Engine

Temperature × Humidity × Traffic = 64 possible mood states. Each mood maps to a curated selection of ambient/experimental music: Harold Budd for cold dry mornings, Jon Hopkins for humid congested afternoons. The system doesn't generate playlists — it makes editorial decisions based on atmospheric context.

Business Model

Proof-of-concept for "The Sensory Stack" — productized automation infrastructure designed to be packaged and sold. System validates that atmospheric data can drive consistent daily content generation without manual intervention. Currently demonstrates technical feasibility over 6 months of operation.

02 Live — Production

OfferFixer

Hormozi $100M Offers framework automation

offerfixer.cdx.la

The Problem

Founders and freelancers don't know how to structure offers. They default to hourly rates or vague "I do X" positioning because applying frameworks like Hormozi's $100M Offers takes hours of iteration most people abandon.

The Solution

Free tool that takes user inputs (product, price, client, result) and returns structured offer diagnosis using Hormozi's Value Equation: Dream Outcome × Perceived Likelihood / Time Delay × Effort. Includes pricing guidance, copy ready to use, and 5-step action checklist.

9
Offers Processed
6
n8n Nodes
1 Day
Build Time
Frontend
Cloudflare Pages
Backend
n8n webhook
LLM
GPT-4o-mini (OpenRouter)
Storage
PostgreSQL logging
Deployment
offerfixer.cdx.la
Launch
April 2026

Workflow

Webhook Build Payload OpenRouter Parse Response Postgres Log Respond JSON form LLM JSON analytics output

Prompt Engineering

The system prompt applies Hormozi's Value Equation systematically: diagnoses which variable is weakest (Dream Outcome, Perceived Likelihood, Time Delay, or Effort), then rebuilds the offer addressing that specific gap. Temperature set to 0.25 for consistent framework application, max 700 tokens to force concise output.

Business Model

Free tool, no signup required. Built in one day to validate LLM application of Hormozi's framework. Demonstrates ability to encode strategic frameworks into executable prompts with structured output. Currently processing organic traffic.

03 Live — Production

Track Vibe Generator

Real-time AI descriptions for streaming overlays

The Problem

Streamers and DJs want dynamic overlays showing "now playing" with context, not just artist + title. Manual descriptions don't scale when you're playing 30+ tracks per session, and static metadata is boring.

The Solution

Webhook-based system that receives track metadata from Spotify or Last.fm, generates contextual descriptions via OpenAI in Spanish + English, and returns JSON for frontend display. Integrates with OBS overlays for live streams and Club Sensorial broadcast sessions.

10
Live Executions
4
n8n Nodes
2 Lang
ES + EN Output
Backend
n8n webhook
LLM
OpenAI
Input
Spotify/Last.fm APIs
Output
JSON (text + text_en)
Frontend
HTML/JS overlays
Use Case
OBS streaming

Workflow

Frontend Poll n8n Webhook OpenAI Generate Code Format Respond JSON track prompt vibe ES+EN

Frontend Integration

Two HTML overlay variants: Spotify (OAuth refresh token) and Tidal/Last.fm (public API polling). Both poll every 2-3 seconds for "now playing" changes. When track changes, POST to webhook with artist + title. Response displays below album art with fade-in animation. Designed for OBS browser source with transparent background.

Prompt Engineering

OpenAI prompt generates 1-2 sentence contextual descriptions in Spanish, followed by English translation. Focus on mood, sonic characteristics, and cultural context rather than generic metadata. Temperature set for creative but consistent output. Max tokens limited to keep overlay text concise for streaming display.

Use Case

Built for Club Sensorial streaming sessions: ambient/experimental music broadcast with AI-generated context. System runs during live sessions — currently 10 executions logged, last used 2 days ago. Simple architecture (4 nodes) but production-stable for real-time use.

04 Architecture — Not Deployed

ZEITBRIEF Pipeline

Complete automation architecture (pivoted to manual MVP)

zeitbrief.cdx.la

Original Problem

Builders need daily signal filtering from 1,000+ tech/startup/AI sources. Generic newsletters don't answer "what can I build with this?" Manual aggregation and synthesis doesn't scale. Content fatigue when every brief is 10 paragraphs nobody reads.

Technical Solution (Built)

Complete 27-node automation pipeline: Tavily API aggregates 50+ sources → RSS feeds supplement → OpenRouter synthesis with strict 3-block editorial structure (SIGNAL / SO WHAT / BUILD THIS) → Slack human approval gate → Beehiiv publishing → PostgreSQL logging → landing page auto-update. System validates technically but was not deployed to production.

27
n8n Nodes Built
6
Workflows Designed
Manual
Deployed as MVP
Automation
n8n (2 workflows)
LLM
Sonnet (OpenRouter)
Sources
Tavily API
Storage
Postgres (zeitbrief_outputs)
Approval
Slack HITL gate
Publishing
Beehiiv API

Workflow

Manual Trigger Read Postgres Parse Payload Slack Approval Mark Published Beehiiv Publish init draft content DB email

Editorial Structure

Every brief follows strict 3-block structure: SIGNAL states only facts (what happened, no opinion), SO WHAT explains implications for people who build things (why it matters to you specifically), BUILD THIS proposes concrete projects you could ship this week. No fluff, no 10-paragraph analysis.

Why Not Deployed

System architecture complete and technically validated. Decision to pivot to manual MVP (Codex + local repo) for faster content iteration during market validation. The bottleneck wasn't automation — it was finding product-market fit for the editorial angle and buyer targeting. Manual generation allows daily experimentation with offer framing, niche selection, and tone without re-engineering prompts in n8n.

Current Deployment

ZEITBRIEF now runs as ZeitWise: manual pipeline via Codex with structured repo (runs/YYYY-MM-DD/, JSON schemas, metrics ledger). Published to LinkedIn newsletter (90 subscribers). Full n8n automation exists as reference architecture for future deployment once content-market fit validates.