The problem
You have data: sensor logs, ERP dumps, drawings, tickets. A board that wants "something with AI". The big consultancies bill slides; the freelancers stop at notebooks. Nothing ships, and whatever does ships straight into a US cloud.
Boutique AI consultancy · EU
Your data. Your infrastructure. Your AI. We help European SMEs adopt AI without surrendering their data, their IT, or their future to US hyperscalers. Fixed-price entry. Numbers on the page. A runbook that lets you walk away from us if you ever need to.
01 · Why us
You have data: sensor logs, ERP dumps, drawings, tickets. A board that wants "something with AI". The big consultancies bill slides; the freelancers stop at notebooks. Nothing ships, and whatever does ships straight into a US cloud.
GDPR, the EU AI Act, and the wider push for European digital sovereignty all pull the same way: European firms want their data and their AI under European law, on European infrastructure, with an exit plan.
A boutique that is senior, fast, fixed-price on entry, and sovereign by default. We scope in two weeks, deliver a tailored build, and hand you the keys: weights, infra, runbook. Including how to leave us.
YOUR DATA. · YOUR INFRASTRUCTURE. · YOUR AI.
The three things every engagement leaves you in charge of.
02 · Services
We package our work into two commercial themes. Every sub-service is sovereign by default: EU-hosted, on-prem-capable, audit-trail-ready.
Drive out downtime, defects, and waste.
Own your data, your infrastructure, and your AI.
03 · Specialisation
Most "sovereign AI" vendors stop at a Llama in a container. We go end to end. Four layers, one engagement, one runbook.
EU-only data flows, mapped against GDPR Art. 46 and Schrems II. Minimisation, retention, and purpose-limitation audits. Sovereign embeddings and vector stores. Audit-ready lineage.
European hosting (Hetzner, OVHcloud, Scaleway, StackIT, IONOS, Leaseweb), or on-prem when the client needs it. Open-source default stack. No US-SaaS lock-in. Every engagement ships a "how to leave us" runbook.
On-prem and EU-cloud LLMs, led by Mistral and other open-source models (Llama, Phi, EuroLLM), with privacy-first RAG. Domain-adapted models with the weights in your account. Agentic workflows with full audit trails and EU-AI-Act-grade risk gates.
Monitoring, evals, incident playbooks, DPO-ready documentation, and staff training so the system lives after we leave.
04 · Approach
A focused workshop, followed by a written report with options you can take to your board and your auditor. Two weeks. One price. One document.
Entry product
EUR 4,950
fixed · ex VAT · two weeks end-to-end
A workshop and a written report. Deliberately sized so you can say yes this quarter, without a procurement detour.
A half-day working session with your ops and IT leads. We map processes, the data you actually have, and the vendors and jurisdictions it lives under today.
A GDPR and EU AI Act reality-check for your specific setup. No theatre: a vendor map and a red / amber / green assessment.
One or two high-impact, technically feasible, sovereign AI use-cases. Each scored on value, feasibility, and sovereignty risk.
A written report with the options on the table, trade-offs called out, and a recommended next step. Your name on the cover. Board-ready.
We scope a tailored next project against the recommendation in your report. No off-the-shelf retainers, no pre-priced pilots. Each engagement is sized to the roadmap your report recommends and to the pace you want to move.
05 · Proof
Theme 01 · Operational
EUR 50kdirect cost savings
Statistical time-series models inside a digital-twin framework, with a PowerBI layer for non-technical stakeholders. Research translated into measurable business value for an industrial SME.
Theme 02 · Sovereign
0 bytesleave the device
A fully sovereign, locally deployable meeting-notes system. Privacy-first LLMs, no external calls, no US cloud surface. Built in an applied-AI research context.
Theme 02 · Sovereign
GDPR-nativeprivacy-first legal AI
Domain-specific LLM and RAG architecture for mediation under Dutch law. Privacy-compliant, on-prem-ready, designed around a regulated workflow.
Theme 01 · Operational
+5%MAPE vs. SOTA
A time-series foundation model fine-tuned for building-level electricity demand, benchmarked against LSTM+CNN and Prophet under extreme weather anomalies.
06 · What we build
After the Scan, the shape of the next project is driven by the use-case. We design and deliver across these forms, always sovereign by default.
Custom internal tools, operational dashboards, and board-ready reporting surfaces. EU-hosted, SSO-ready.
Secure, branded portals for document exchange, workflow sign-off, and customer self-service.
On-prem or EU-cloud LLM assistants over your data. Mistral and other open-source models, privacy-first RAG.
Forecasting, anomaly detection, and classification on time-series, sensor, and operational data.
On-prem or edge CV for inspection, defect detection, and remote monitoring.
Ingestion, vector stores, evaluation harnesses, and the observability every production AI system needs.
07 · Team
We speak both your languages: deep technical AI and practical European business. And we build so you can walk away from us if you ever need to.
Co-founder · Business & Communications
Marketing and communications lead with deep European SME experience. Owns the client relationship, positioning, brand voice, and the translation of technical output into board-ready business cases. Fluent in the Dutch commercial register.
Co-founder · AI & Research
Senior AI researcher. PhD in Applied Mathematics (TU Delft / Deltares). 10+ years building real-world industrial AI. Production work across time-series forecasting, computer vision, GeoAI, and sovereign LLMs.
08 · Let's talk
Book a free 30-minute discovery call. If it's a fit, we scope a Sovereignty & AI Scan the same week.