luisdelatorre.de
Cloud & AI Solutions Architect, Hamburg
I design and build agentic AI systems, on AWS or where the data has to stay in Europe, and I teach engineering teams to work with them.
Agentic AI systems from architecture through to production. I have built several of them, on AWS and on EU sovereign infrastructure. That means I can deliver on a hyperscaler, and I can deliver when the data has to stay under European control.
A two day workshop for engineering teams, followed by optional mentoring.
German companies with fewer than 50 employees can have up to 100 percent of the training cost subsidised through the Qualifizierungschancengesetz.
An engineering office that reviews railway construction submissions wanted to know whether an AI agent could check real plan sets against real regulations without inventing findings. In five weeks I built a browser-based system that checks completeness, text requirements and the drawings themselves. The drawings were the focus: that is the part a text-only check cannot reach. Every finding carries its regulation reference, document, page and quote, or it is not reported.
A municipal training provider wants to bring AI into its courses without giving up control over where its data goes. We are building an MVP of a learning platform on one non-negotiable: model, data and hosting stay in Europe. The architecture is deliberately small. Every AI step sits behind a human decision, and the model runs behind an abstraction layer so it can be swapped without touching the rest of the system. Where a common pattern is not needed yet, we leave it out and document the trigger that would bring it back. Sensitive data is kept away from the model by design, not by policy. The goal is an MVP the client can run, extend and audit from day one, not a demo that has to be rebuilt before production.
Two two-day workshops with the engineering organisation of a German insurance broker pool, then five weeks of mentoring with three of its teams. The workshops run the four pillars and a guided lab that builds an application from an empty repository, with architecture guidelines and hooks that enforce them, then a second day on real issues from the teams' own code. The mentoring is the part that decides whether any of it survives contact with a backlog: the same cycle, run by the teams themselves, on their own work, week after week.
Let us talk about your project or your team.
I work remotely from Hamburg and travel for occasional onsite days across Europe.