Senior Solution Architect · AI/ML · Sovereign Cloud

Wojciech Nakielski

I build AI systems that belong to the company running them.

My subject is how organisational knowledge gets into AI systems without leaving the building. At NOVEDAS Consulting (COHEMI Group) I lead the AI focus group. My day-to-day work is cloud and interface architecture in the automotive sector. What I build in AI in my own time sits on GitHub, together with the architecture decisions behind it.

Portrait of Wojciech Nakielski
22 yearsin IT, starting out as a software developer
Solution ArchitectCloud and interface architecture in the automotive sector
AI focus groupLeading the internal AI strategy at NOVEDAS
Focus areas

What I work on

Second Brain · RAG · MCP

Getting knowledge into AI systems

The model is rarely what fails. The bottleneck is access to your own knowledge. I build systems where an agent can find what it needs and show why. Similarity scores alone will not do that, and the knowledge stays in the building.

STACKIT · Local LLMs · GDPR

Sovereign AI & data protection

AI whose data stays in-house: local models, European infrastructure. GDPR and the EU AI Act go into the design from the start.

Agents · Orchestration · MCP

AI architecture & agents

Agentic systems that make it from prototype into production. Part of that is being able to reconstruct later why a system decided the way it did.

AWS · Microsoft · Google · arc42

Cloud, software & systems architecture

Scalable cloud architectures, interface and database design, documented to arc42. I have been doing this since 2004, among other things in the automotive sector.

Current article

How an agent finds what it already knows

raw/

Source documents and raw data, immutable. The agent only reads here.

wiki/

Knowledge the agent generates itself: insights, concept links, indexes.

system/

Contracts, prompts, and templates for operation.

Build in Public · 4 min read

Why My AI Agent Has a Second Brain

A file-based knowledge system for autonomous agents. Self-hosted, no vector database.

Read the article →

View all articles →

Career

22 years in IT, from software developer to solution architect

since 2023

Solution Architect / IT Architect

NOVEDAS Consulting GmbH (COHEMI Group)
2020–2022

Senior IT Consultant

ISR Information & Products AG

View full profile →

Build in public

Projects

Alongside consulting I build my own software. Private projects, documented in public. Anyone can read up in the repos on how the architecture decisions came about.

All projects on GitHub →

Contact

How to reach me

AI systems and the architecture underneath are my subject. I am happy to talk about them any time, easiest by email or on LinkedIn.

What I enjoy talking about

AI & Agents

  • LLMs & agent systems
  • RAG & embedding strategies
  • Model Context Protocol (MCP)
  • Local models (Ollama, LM Studio)
  • Agentic engineering
  • n8n · Langflow

Sovereign AI & Governance

  • EU AI Act · GDPR
  • AI governance & policies
  • STACKIT (sovereign cloud)
  • On-premise & self-hosting
  • Confidential computing

Data & Knowledge Systems

  • Vector databases (Qdrant, pgvector)
  • Embedding & chunking strategies
  • PostgreSQL · MySQL · Oracle
  • Neo4J · ElasticSearch
  • Schema design & modelling
  • Query & performance tuning

Cloud & Infrastructure

  • AWS (CloudFormation, CDK)
  • Microsoft 365 Copilot & Copilot Studio
  • Google Vertex AI & Gemini Enterprise
  • Kubernetes & containers
  • Infrastructure as code

Architecture & Method

  • Enterprise & solution architecture
  • Interfaces & integration
  • Migration strategies
  • arc42 · UML · BPMN
  • Requirements engineering
  • SCRUM & quality assurance

Industries

  • Automotive
  • Banking
  • Aviation
  • Public sector
  • Energy
  • Wholesale
  • Gaming