Skip to content
Noel Barral
ES
Menu

Available for senior AI & automation roles (remote)

Noel Barral Senior AI Solutions & Automation Engineer

Applied AI systems architecture, deterministic automation pipelines and resilient production infrastructure. Over 20 years bridging systems engineering and business bottleneck resolution.

Download technical CV (PDF) Direct contact · [email protected]
Noel Barral
  • Python
  • TypeScript
  • n8n
  • Supabase
  • PostgreSQL / pgvector
  • Qdrant
  • Docker
  • Linux Debian
  • Nginx
  • Git
  • Tailscale
  • Cloudflare
  • Redis
  • Firecrawl
  • Zammad
  • Mistral OCR
  • Gemini Vision
  • DeepSeek
  • JEV (TypeSafe)
Systems engineering & infrastructure
20+ years
Dedicated Linux servers in production
11+ years
Deployed architectures, described below
4

Engineering Pillars

Three pillars, one stack

I don't collect technologies. I design systems where the model, the pipeline and the host that runs them are planned together.

01

Applied AI & Knowledge Retrieval

  • Design and delivery of production-grade RAG (Retrieval-Augmented Generation) architectures.
  • Knowledge bases decoupled into layers (static, contextual, dynamic) on top of vector stores (pgvector, Qdrant).
  • Human-in-the-loop review workflows to mitigate hallucinations and keep context evolving.

02

Process Automation & Data Pipelines

  • Orchestration of mission-critical workflows with n8n, Python scripts and RESTful API consumption and exposure.
  • High-volume extraction pipelines (distributed web scraping), metric normalisation and JSON/XML structuring.
  • Event-driven automation through asynchronous webhooks and message queues.

03

Platform Infrastructure & DevSecOps

  • Administration and hardening of dedicated Linux environments (Debian/Ubuntu), built up over more than a decade.
  • Containerised architectures on Docker and secure peer-to-peer interconnect through a Tailscale mesh.
  • Edge protection with Cloudflare (WAF, DNS, DDoS) and continuous observability and audit with Wazuh SIEM and Grafana.

Case Studies

Architectures in production

Four systems running in production right now. These are not exercises: each one removed a measurable bottleneck.

Case 01

RAG system & multi-channel operational triage

Support and incident-resolution platform for a low-cost messaging operator, covering written and telephone channels.

The problem

Support channels saturated by repetitive tracking, quoting and logistics queries, with high contextual variability, both in writing and over the phone.

The technical solution

  • Help desk deployed on Zammad, wired two-way through webhooks to an n8n orchestration engine, with Supabase (PostgreSQL) holding cases, conversations and triage state.
  • Hybrid RAG architecture: dense vector search combined with heuristic rules for direct triage.
  • Text and data extraction from images, delivery notes and invoices through vision pipelines: Mistral OCR for documents and Gemini for visual understanding, so triage has the content even when the customer only sends a photo.
  • Voice handling: a system that watches the phone voicemail boxes, transcribes the messages, classifies them with the same triage engine and replies automatically when the case does not require a human operator.
  • Human-in-the-loop protocol: cases scoring below the confidence threshold are routed to human operators with pre-generated drafts, and every validation feeds back into the vector store.
  • Stack:
  • n8n
  • Supabase
  • Zammad API
  • pgvector
  • Mistral OCR
  • Gemini Vision
  • Docker
  • Python

Case 02

Reverse catalogue ingestion & visual refurbishment generation

Automated technical quoting for CID Construcciones.

The problem

Construction suppliers' technical catalogues scattered and inconsistent, non-standard dimensions, and slow turnaround on visual quotes for customers.

The technical solution

  • Scraping pipelines built for systematic capture of images, dimensions and technical datasheets at scale.
  • Metric normalisation and metadata ingestion into relational database schemas.
  • Wired to generative vision models to project modifications onto floor plans and produce automated technical quotes.
  • Stack:
  • Python
  • BeautifulSoup / Playwright
  • Next.js
  • REST APIs
  • Computer Vision

Case 03

Unified lead ingestion & heuristic spam filtering

Lead-capture infrastructure for Mudanzas Go.

The problem

Incoming requests fragmented across several standalone landing pages, with operations swamped by junk leads and automated spam.

The technical solution

  • Asynchronous webhooks centralised in n8n.
  • Intermediate deterministic anti-spam layer with entity validation (phone, postcode, load volume).
  • Integrated with the corporate knowledge base to fire assisted operational replies and sync with the CRM in real time.
  • Stack:
  • n8n
  • Cloudflare Workers/WAF
  • Webhooks
  • OpenAI API

Case 04

B2B industrial ontology & compatibility graph Stealth mode

Industrial SaaS platform in development.

The problem

OEM versus aftermarket part numbers extremely fragmented, causing mis-purchases and disconnected catalogues.

The technical solution

  • Ontology model plus a relational/vector database tuned for cross-reference mapping.
  • High-volume ingestion pipeline with semantic resolution of equivalences across manufacturers.
  • Semantic search engine tolerant of typos and workshop jargon.
  • Stack:
  • FastAPI / Python
  • PostgreSQL
  • Qdrant
  • Docker
  • Tailscale

The Trajectory

From bare metal to AI systems

Modern solution engineering doesn't stand on a vacuum of cloud libraries. My technical track began over two decades ago in the United States, building LANs and configuring Windows NT and Mac servers at Montachusett Regional Vocational Technical School (Monty Tech).

That physical foundation was consolidated working directly on critical infrastructure and data switching in the ONO and Orange data centres, and running dedicated Linux production servers for more than 11 years at DoctorSEO. When I design a solution with LLMs or automate a workflow in n8n, I do not just understand the API call: I understand container memory, the encrypted network tunnel, port security and whether the business model holds.

  1. 2002 — 2003 Monty Tech · IT Shop

    LANs, Windows NT and Mac. The physical foundation for everything that came after.

  2. Data centres ONO / Orange

    Critical infrastructure and in-room data switching.

  3. 11+ years DoctorSEO · Linux in production

    Dedicated server administration, hardening and service continuity.

  4. Present Applied AI & automation

    RAG architectures, n8n orchestration and containerised platforms in production.

Bonus

Away from the keyboard

Skiing, motorcycling, photography, cycling, Magic: The Gathering and offensive security: what I explain in six photos.

See my hobbies

Contact

Let's talk about systems

Open to remote contracts and senior engineering roles (Spain / EU).

Email [email protected] LinkedIn linkedin.com/in/noel-barral CV Technical CV (PDF)