I design and build AI systems from scratch: from multi-agent architecture and knowledge graphs to the infrastructure that runs them. For organisations that need a solution built to their requirements, not another pilot with an off-the-shelf tool.
Why custom architecture
Off-the-shelf AI tools cover a lot, but not everything. When the problem is specific, the data sensitive and the integration with the process demands precision, you need an architecture designed for the exact requirement. I build such solutions myself — and on them I test approaches that I then apply in work for organisations.
What I build
- Multi-agent systems for regulatory compliance automation: screening, risk classification, document analysis.
- RAG (Retrieval-Augmented Generation) architectures connecting knowledge bases with business processes.
- Knowledge graphs with vector databases: relationships between entities, documents and regulations.
- Clinical decision-support systems for healthcare: rare diseases, clinical documentation analysis.
- Tools for secure on-premise data processing (data sovereignty): anonymisation, locally run language models.
- AI infrastructure: my own GPU server with dozens of containers, vector databases, orchestration engines and language models.
How it works in practice
We start from the problem: what the system should do, on which data, with what level of human oversight and where it should run (cloud, on-premise, hybrid). I design the architecture, build a prototype, test with users and take it to production. The whole can run on the client’s infrastructure or on mine — depending on data and security requirements.
Who it is for
Organisations that have a specific problem to solve with AI and need someone who will design and build the solution, not only advise. Also teams that want to see what a production AI architecture looks like from the inside.
Why me
I build and maintain my own AI systems on my own GPU infrastructure: dozens of containers, vector databases, orchestration engines, document processing and language models. I combine that with twenty years of designing processes and auditing management systems, so the solution is not only technical but built into the organisation’s process and governance.
Let’s talk about solution architecture
See also: AI audit, AI workshops, AI governance and ISO 42001 and fractional Chief AI Officer.