About KAIX LAB
Who we are.
We are an independent AI engineering and cybersecurity lab. We build systems that go into production and audit the ones already running. Ten years in the craft.
01 — Our story
Where we come from.
KAIX LAB was born inside real AI projects: agents leaking data, RAG systems with no access control, models in production with no evaluation evidence. We saw too many prototypes that collapsed the moment real users showed up.
We decided to build a lab that combines two disciplines the industry tends to keep apart: AI systems engineering and offensive security. People who build agents with security by design audit better; people who red team LLMs design sturdier architectures.
We work with European companies: SaaS startups, clinics, law firms, customer-service platforms.
02 — Principles
How we decide what to ship.
These four principles define which projects we accept, how we run them and what we put in writing. They are not slogans: they are the filters we apply every week.
- 01
Bounded scope
Every proposal names a use case, a scope and concrete deliverables. No open-ended projects.
- 02
Production quality from day one
Evaluation, observability and security controls from the start of the project, not bolted on at the end.
- 03
We say what we think
If AI isn't the right answer for your case, we tell you. If a simpler model is enough, the same.
- 04
European frameworks
We work with EU AI Act, GDPR, NIS2 and ISO 42001. We know them in detail and apply them to your case.
03 — How we work
Four phases, one responsibility.
Every engagement follows the same structure, sized to scope. No surprises in the invoice or the calendar: deliverables and acceptance criteria are agreed in writing before we start.
Diagnosis
Technical session to understand the system, the data, the regulatory risks and the real goals. Output: signed scope and plan.
Construction
Short iterations with continuous evaluation. Code in your repository, technical documentation and weekly demos with your team.
Hardening
Red teaming, prompt injection, access control and evidence review. We close findings before the production rollout.
Operation
Hand-off to the internal team with runbooks, metrics and a maintenance plan. Optional ongoing support with an agreed SLA.
04 — In numbers
What backs up what we say.
Conservative, verifiable figures. We'd rather not count a project than invent a number.
- Experience
- 10+ years
- Practices
- 2
- Regulation
- EU AI Act
- Geography
- EU
In applied AI, big data and offensive security in production.
AI & Automation and AI Cybersecurity & Compliance, integrated.
GDPR, NIS2 and ISO 42001 covered when the system requires it.
Based in Madrid, on-site with clients across Europe when the project demands it.
05 — Team
You talk to whoever does the work.
The person on your first call is the one who will write the code or sign the audit. You always talk to whoever is doing the work.
When a project needs more hands, we bring in trusted collaborators with a verified technical profile.
Let's talk about your AI system.
We respond within 24 – 48 business hours. We'll suggest a first call to understand your case.
[email protected]