Software, business and AI consulting

Technical judgement to decide what to do before you build.

I analyse your problem and help you decide what to do, how to do it and where to start.

See how I work ↓

01Services

What I can do for you

Four concrete problems where business and technology need to move forward together.

Not sure your case fits? Tell me in two lines.

02Method

How I work

Five steps, and you don't have to climb them all.

first session

The first conversation

Understanding the problem and scoping it. No deliverables.

fixed price

The map

Your knowledge, ordered and navigable.

hourly

Judgement

What is dead weight, what to redirect, what will get expensive.

scoped

Test

A prototype that proves or rules out that something fits.

budget per phase

Implementation

In phases, each with its own scope and price.

Pricing agreed in the first conversation · 50% up front, 50% on delivery · no maintenance.

The first step is half an hour of conversation. No strings attached.

03Cases

What I have built

Discover my products, open-source projects, client work and research.

Products

zetesis rag — Búsqueda y agentes sobre tu propia documentación

graphmarks — Tus marcadores, como un mapa

sebastian — Voz bidireccional sobre hardware propio: el altavoz sobre la mesa, con el anillo de luz encendido

Projects

Konect

Konect: audio transcription and analysis

ProductKonectTranscription

Homelab

Building a homelab from scratch: the big picture - Homelab (01/06)

Case studyHomelabProxmox

Escohotado

Antonio Escohotado's legacy, explorable with AI

Case studyEscohotadoRAG

CERN's IT governance graph, with its nodes and relations

CERN graph

CERN's IT governance, navigable as a knowledge graph.

CERNQuartzKnowledge graph

Open the graph

CERNphone

CERNphone: telephony for more than 5,000 users

ProductCERNphoneTelephony

Keycloak

Centralized identity management with Keycloak

LabKeycloakRBAC

Naiz Fit

Naiz Fit: configurable testing campaigns

Case studyNaiz FitProduct

Zetesis Portal

Zetesis Portal and PayloadAgents: how we built our SaaS

ProductZetesis PortalPayloadAgents

El panel del optimizador fiscal foral: el ahorro identificado del ejercicio en grande, la barra del crédito vitalicio de vivienda agotado a 36.000 €, y el desglose por vivienda, EPSV e inversiones, con el pie que recuerda que todo es una proyección del mismo log de eventos y que la herramienta es orientativa.

Simulador de la Renta

A simulator for Bizkaia's foral income tax

LabEvent sourcingTaxation

Las dos mitades del montaje, separadas por una diagonal. A la izquierda, sobre fondo oscuro, la gráfica de coste por modelo de Langfuse a lo largo de tres meses, con el gasto acumulado del periodo. A la derecha, la tabla de la pasarela: cuatro alias estables —chat-estandar, chat-premium, razonador y economico— apuntando a modelos de cuatro proveedores distintos, cada uno con su coste de entrada y de salida por millón de tokens.

LLMOps

LLMOps on your own cluster with LiteLLM, Langfuse, and ArgoCD

LabLLMOpsGitOps

Postiz

Postiz: scheduling social media posts from your own server

LabHomelabKubernetes

Harbor

A Harbor of our own for the cluster's images and charts

LabHarborGitOps

Go to the blog →

Collaborations

Facing a similar problem?

04About me

Rubén García Hernando

Software architect and researcher

I could not make sense of this work without research. After three years at CERN, back in Bilbao I decided to work on my own and keep researching and learning. I believe entrepreneurship is the ultimate professional challenge — and, at the same time, an expression of creativity and a claim to freedom.

Zetesis comes from the Greek ζήτησις, inquiry: find the cause before proposing the solution.

Learn more

2023 — today

Zetesis

Founder. My own platform, in production.

2023 — 2024

Naiz Fit

iOS and Android apps used at scale.

2019 — 2022

CERN

Internal telephony: 30,000 calls a day.

2017 — 2019

LIN3S

From developer to software architect.

Software development and AI

Standardise how your team builds software with AI

The problem

Your team already builds software with AI, but everyone works differently. Code arrives faster, while inconsistency, technical debt and the cost of reviewing it all keep growing.

How I help

I define a shared method with your team for giving AI the right context, reviewing its output and measuring whether it genuinely improves delivery.

AI amplifies the human factor

The tool multiplies each developer’s capacity, but the result still depends on how they understand the problem, how well they know the system and which safeguards they apply. Two people using the same model and repository can produce radically different outcomes.

I therefore treat adoption as a matter of personotecnia, inspired by Javier G. Recuenco’s “X factor”. People’s capabilities, motivations and ways of working are not peripheral to the problem; they are part of it.

Standardisation does not mean making everyone code in the same way. It means creating a system where different strengths add value without putting the product at risk.

Separate exploration from product work

A spike exists to reduce uncertainty. It can move quickly, test alternatives and discard code. Product software lives under different constraints: it must fit what already exists, remain understandable and keep working through many future changes.

Separating those paths gives curious developers room to explore without quietly bypassing the rules, and stops experimental code reaching production as though it were ready to sustain it.

Define which work is research and which work changes the product.
Agree what must be recorded when a spike ends.
Create an explicit step for turning learning into maintainable software.

A shared method, not a prompt collection

I do not force one tool or prescribe every conversation with an agent. I help the team agree on what must be shared: where context lives, which changes need a specification, who validates architectural fit and which checks protect the product.

Spec-driven development helps when a task contains decisions that should not be improvised. Tests turn part of that intent into an executable guarantee, and continuous integration makes sure every proposed change runs the agreed checks.

Reusable repository context and instructions.
Clear criteria for when a specification is needed.
Behaviour-focused tests and short feedback loops.
Required checks on pull requests.
Skills that capture methods the team has already proved useful.

Adoption and support

I begin by observing how the team actually works: what it delegates, where errors appear and who has already developed practices worth sharing. We then distinguish types of work and apply a level of discipline proportionate to each one.

Adoption includes hands-on training, support through the first changes and measurement of lead time, defects, review effort and maintainability. The aim is not more code. It is better software without losing control of how it is built.

What changes

A team capability you can repeat and improve

01A shared method for working with coding agents.
02Different guardrails for research, product work and critical systems.
03Specifications and tests that preserve the intent of a change.
04Continuous integration that applies the agreed controls.
05Practical training on the team’s own code and problems.

Go deeper

Specifications, tests and CI: setting boundaries for coding agents Boris Cherny’s five archetypes

Tell me how your team is using AI

A first conversation can reveal where quality is being lost, where a valuable practice already exists and which change would have the greatest impact.

Software, business and AI consulting

What I can do for you

Standardise how your team builds software with AI

Put your company knowledge to work with AI

Turn your prototype into a sellable product

Unblock problems caught between business and technology

Standardise how your team builds software with AI

Put your company knowledge to work with AI

Turn your prototype into a sellable product

Unblock problems caught between business and technology

Standardise how your team builds software with AI

Put your company knowledge to work with AI

Turn your prototype into a sellable product

Unblock problems caught between business and technology

How I work

The first conversation

The map

Judgement

Test

Implementation

What I have built

Products

Zetesis RAG

Graphmarks

Sebastian

Projects

Konect

Homelab

Escohotado

CERN graph

CERNphone

Keycloak

Naiz Fit

Zetesis Portal

Simulador de la Renta

LLMOps

Postiz

Harbor

Collaborations

Rubén García Hernando

Zetesis

Naiz Fit

CERN

LIN3S