Start here

The three verticals

The blog is organised into three training verticals aligned with the itineraries of lo0 (training) and with the projects of Fibercli (sovereign infrastructure and cybersecurity consultancy):

  • AI — LLMOps in production, LLM inference, data and RAG, training, evals, guardrails and observability.
  • Kubernetes — lightweight distributions (RKE2), eBPF networking with Cilium, observability with Hubble, runtime security with Tetragon.
  • Networking — Huawei EVPN/VXLAN, MikroTik RouterOS, Ubiquiti UniFi, HCIA Datacom certification.

The three verticals share a common denominator: technical content aimed at real operation, not at marketing.

Every article on the blog is published in Spanish and in English. The language selector in the header switches between the two versions of whatever page you are on.

Where to start, by profile

If you are here for the AI vertical

If you have never worked with LLMs in production, two posts, in order:

  1. MLOps for LLMs in 2026 — the landscape.
  2. The six-stage LLMOps pipeline — the master map.

If you already operate an LLM system and want to go deeper:

If you are weighing OSS against cloud:

If you are here for the Kubernetes vertical

If you are interested in how LLMs are served on K8s, the posts vLLM on Kubernetes, LLM operators on K8s and Multi-tenant GPU cluster live in the AI vertical but carry the kubernetes tag for cross-visibility.

If you are here for the Networking vertical

The canonical post for each vertical

VerticalRecommended post
AIAnatomy of an LLM request in production
KubernetesRKE2 with Cilium BGP
NetworkingEVPN/VXLAN on Huawei

If you would rather see the whole picture at once, the blog map orders everything published into the ten layers that separate the utility feed from the token served to a user.

How this blog is built

It is the technical blog of lo0 (training) and Fibercli (sovereign infrastructure and cybersecurity consultancy). The editorial rule:

  • No marketing, no hype. Figures come with their source or a derivable calculation.
  • When something is an estimate, it says so.
  • On-premise hardware always in generic market configurations, never specific internal infrastructure.
  • Bidirectional cross-links: the blog is a navigable network, not a chronological sequence.

Subscribe or follow updates