The person behind AnyLAI
Luiz Hogrefe
Luiz Claudio Hogrefe · Software architect · Cologne
Luiz Hogrefe is a software architect based in Cologne. His work sits where software architecture meets governed AI, semantic interoperability and traceable digital evidence, including the architecture of digital product passports.
AnyLAI is where this work is developed and published in the open: a framework, a working demonstration and regulatory writing that a technically serious reader can open, check and disagree with.
At a glance
- Full name
- Luiz Claudio Hogrefe
- Role
- Software architect
- Based in
- Cologne, Germany
- Building software since
- 2008
- Working languages
- Portuguese, German, English, Spanish
- Profile
What he works on
The common thread is what he calls context management: making sure that meaning survives when data moves across systems, organisations, languages and jurisdictions.
Software and solution architecture
System boundaries, integration models and technical roadmaps, written down so that they can be reviewed, tested and handed over.
Distributed and cloud-native systems
Java and Spring backends, event-driven integration, Kubernetes, and the delivery pipelines that keep all of it releasable.
Semantic interoperability
Shared vocabularies and data models such as ECLASS and the Asset Administration Shell, so that a field means the same thing at both ends of an exchange.
Governed AI
Architectures in which probabilistic models do bounded work while deterministic rules decide, keep a record and can be replayed.
Traceable evidence and product passports
Provenance that travels with the data, verifiable credentials, and digital product passports that can be checked without trusting the issuer's website.
Regulation as an engineering input
Reading texts such as the ESPR, the EUDR and the AI Act as requirements a system has to meet, not as labels to put on it.
The question behind the work
One question runs through much of this work: what happens to software architecture when implementation itself becomes increasingly automated?
The aim is not to remove people from software development. It is to understand which parts of human engineering judgment have to become explicit, testable and enforceable if AI is allowed to do more of the implementation: what a system is for, which states it must never reach, what its data means, and what evidence it owes the people who rely on it.
The argument in fullLess human coding, more human architecture
Why AnyLAI exists
Architecture is easy to describe and hard to check. A diagram in a slide deck or a principle in a report asks the reader to trust whoever wrote it. AnyLAI exists so that this work becomes public, inspectable artifacts rather than an abstract consulting narrative.
It grew out of independent research, from 2024 into 2025, on semantic interoperability, digital product passports and the European rules that will require them. COADF is one of its public outcomes: a framework for building AI-assisted software under explicit governance. AnyDPP is another: a working demonstration of a digital product passport whose evidence verifies against the credential itself. The blog and the Radar read new regulation as architecture requirements.
All of it rests on one idea, a trust architecture: a claim a system makes should carry the evidence behind it, and a reader should be able to check that evidence without asking the system's author.
AnyLAI is an independent research and development project. It is not a company, and nothing on this page is an offer. Project status
Professional background
He has been building software since 2008, first in Brazil and, since 2014, in Germany. Most of those years were spent inside enterprise projects in regulated environments, in banking, industry and education, where an architecture has to survive audits, legacy systems and handovers, not only a first release.
The roles moved from backend development to technical leadership and architecture. What stayed constant is the work of translating complex, often ambiguous requirements into systems that are explicit, testable and maintainable, together with the people who will run them.
Selected stages
2008–2014
Software development in Brazil
Web systems for academic, financial and administrative management at educational and industry institutions in Blumenau. Bachelor of Computer Science, FURB.
2014–2016
Berlin and Düsseldorf
Java development and private-cloud prototypes, then technical lead for the web and mobile products of a small company in the Düsseldorf area.
2017–2022
Enterprise consulting at Sopra Steria
Consultant and technical lead in projects for banking and industrial clients: Java and Spring backends, architecture reviews, structured test frameworks, CI/CD and automated delivery.
2023–2024
Regulated projects at msg
Backend development, test management and agile delivery in a project in a regulated environment in Cologne.
2024–2025
Independent research
Semantic interoperability, digital product passports and the EU rules behind them, among them the ESPR, the ESRS and the AI Act. The architectural blueprint that became AnyLAI.
Since 2025
AnyLAI
The framework, the demonstration and the writing on this site, developed in the open alongside his professional work.
2026
Software architecture at IW Consult
Senior software architect in Cologne: architecture and risk analysis for enterprise applications, cloud prototypes, arc42 documentation, and workshops that turn ambiguous requirements into roadmaps.
How he works
Four habits, stated plainly. They are the practical side of what COADF sets down as a framework.
Architecture over slogans
A system is described by its boundaries, its interfaces and what it refuses to do. A claim that cannot be traced to one of those is not yet architecture.
Deterministic boundaries around probabilistic parts
Language models are useful and fallible. They do bounded work behind rules that decide, keep a record and can be replayed, never in the place where the decision is taken.
Evidence someone else can check
A result is worth what a reader can verify without trusting its author. Its sources, its version and the person responsible for it travel with it.
Standards without lock-in
Public standards first, W3C Verifiable Credentials, UN/CEFACT and GS1 among them, so that what he builds can be verified, replaced or handed over without him.
Selected public work
Where to see the work itself.
- COADFThe Compliance-Oriented AI Development Framework, version 2.2: eight principles for building AI-assisted software under explicit governance.
- Trust architectureConfidence assessed per attribute, human verification by a registered attester, open standards that let anyone check a passport, and processing that stays in the EU.
- AnyDPPThe demonstration: how Brazilian export documents become evidence that a buyer, an auditor or an authority can check.
- The demonstration passportA digital product passport that anyone can open, with a signature that can be checked against the published issuer key.
- Brazil-EU corridorFirst concrete application of the AnyLAI architecture
- From the blogAI governance needs architecture, not another checklistPolicies and checklists describe how an AI system should behave. Whether it does is decided by its architecture: where uncertainty is represented, which boundaries it cannot cross, and what the system keeps as a record.
Profile and contact
Luiz Hogrefe on LinkedInThe professional profile, with the full career history.
Questions about the work, corrections and conversations about the research go through the project's contact form. Open the contact form
Comments on COADF or on an article can also be made in public. Public feedback
Luiz Claudio Hogrefe is the provider of this site under the Impressum and the controller under the privacy policy. He publishes as Luiz Hogrefe, and COADF credits its author as L. C. Hogrefe: one person, accountable under one name in three forms. Impressum