Project Consulting:
Sovereign AI
Book a callUse AI, keep your data.
For public authorities, banks, healthcare and SMEs whose data must not leave the building.
Open models run on your own hardware today, on a laptop as well as in the data centre. We show you how: as a feasibility workshop or project support. Before you invest, we measure on your data what runs well locally.
What can we do for you?
Feasibility workshop
Start small and measure: we take one concrete process from your organisation and build a prototype together with a local model on your hardware. Afterwards you'll know whether the big step is worth it:- Selecting a suitable use case
- Model selection and hardware sizing
- Prototype: a working workflow or RAG pipeline
- Measuring against your data
- Assessment of quality, costs and limits
- Recommendation and roadmap
1 to 2 days
at short notice on request
Project support
We take you from prototype to production. This covers architecture reviews, model evaluation, operations and hardening of your AI infrastructure.In regular office hours we answer questions from your day-to-day project work, within the budget you set.
from 20 hours, over several months if needed
on request, quote within 24 hours
Your advisor
Johannes advises you himself, with the knowledge from his ongoing projects.All workshops since 2016 in the Hall of Fame →
Johannes Hoppe
is a Google Developer Expert (GDE) and works as a freelance software architect, consultant and trainer with a focus on agentic coding, Node.js, .NET and Angular. He is co-author of the standard German-language book on Angular. Alongside the trainings he works as a lead frontend developer in client projects and applies what he teaches there.
Johannes lives in Leimen near Heidelberg.
Bluesky: @johanneshoppe.de
LinkedIn: @johanneshoppe
Frequently asked questions
◢ Who is this offering for?
Organisations that want to use AI but whose data must not end up in third-party clouds: public authorities and administration, banks and insurers, healthcare, critical-infrastructure operators and SMEs whose data protection officers rightly ask questions. For getting started with agentic coding we also recommend Build with AI.
◢ Why work with us?
We build the solutions ourselves, publish our findings openly on the blog and measure results against your use cases.
We demonstrate every recommendation live before you decide.
◢ What does the consulting cost?
We offer the feasibility workshop at a fixed price, project support by effort within a budget you set. We clarify goal and scope with you beforehand, and you receive the quote within 24 hours.
◢ What are the prerequisites?
All engagements are remote by default. On-site sessions in Germany, Austria and Switzerland are available on request.
For workshops with local models, existing hardware is often enough: a model like Gemma 4 (12B) runs on laptops with 16 GB of memory. For project support we need access to your documentation and source control, via VPN if necessary.
◢ Which tasks do local models handle today?
Local models reliably handle classification, summarisation, extraction and internal search today. We show the difference live and measure it against your use cases. Where frontier quality is needed, we show hybrid approaches (non-critical tasks in the cloud, sensitive ones local).
◢ Which industries have you worked in?
Among others, these:
- E-commerce
- Banking & Financial Services
- Software & IT services
- Wholesale and retail
- Media distribution
- Debt and insolvency counselling
- Energy
- Tourism
- Public administration
- Industry
We’ll send you a selection of references on request.
Still have questions?
+49 151 681 222 66 (Johannes Hoppe)
Standard rate to German mobile networksWe'll get back to you within 24 hours.Book a call
Tell us briefly which process you have in mind. We’ll get back to you with suggested times.
- Call30 minutes, free of charge. We clarify use case, data and hardware.
- Quote within 24 hoursFeasibility workshop at a fixed price, project support within the budget you set.
- Prototype on your hardwareAfter one or two days you know what works locally, measured on your data.