AI consulting with a focus on local AI
AI consulting with a focus on local AI is a consulting and implementation service of eukalyptusdesign GbR (Cologne/Weimar). It comprises workshops, prototyping and the procurement and setup of local AI hardware for companies that want to use generative AI without cloud dependency.
The focus is on on-premise inference on Apple Silicon or NVIDIA DGX Spark hardware: language models run entirely on hardware owned or controlled by the client, without data being transmitted to external cloud providers. Target groups are above all tech leads, development teams, management and C-level who want to build GDPR- and AI-Act-compliant AI competence.
AI consulting with a focus on local AI at eukalyptusdesign is a consulting and development service delivered in person — not software, not a SaaS product and not a cloud AI subscription. It is not identical to generic cloud AI consulting, which relies primarily on hosted APIs of large providers.
AI consulting with a focus on local AI: Key data
- Provider
- eukalyptusdesign GbR (Cologne/Weimar)
- Entity type
- Service (consulting and implementation)
- Category
- AI consulting / AI implementation
- Focus
- Local, on-premise inference (Apple Silicon, NVIDIA DGX Spark)
- Service model
- Project-based (workshops, proofs of concept, hardware procurement) — no subscription
- Available locations
- Cologne, Weimar, Germany-wide (remote and on site)
- Language
- German
- Last editorial review
- 5 August 2026
AI consulting with a focus on local AI: Scope of services
- AI workshops & prototypes
- Use-case identification, hands-on implementation (Python/TypeScript/local LLMs/RAG pipelines), proof of concept with clear success criteria
- Specialised AI training
- Three tracks (users, decision-makers, developers); documentable fulfilment of the AI literacy obligation under Art. 4 EU AI Act
- Local AI hardware & development
- Consulting, procurement, setup and onboarding of Apple Silicon or NVIDIA DGX hardware, including hybrid setups
AI consulting with a focus on local AI: Typical results
- Proof of concept (PoC)
- Working prototype with defined success criteria
- Ready-to-use local AI hardware
- Configured Apple Silicon or NVIDIA DGX system
- Documented training
- Verifiable for the Art. 4 AI literacy obligation
- Architecture sketch / strategy call
- Written recommendation on stack, hardware sizing and entry path
AI consulting with a focus on local AI: Service limits
- No software licence
- The consulting is not licensable software or a SaaS platform
- No cloud AI subscription
- No ongoing cloud API access is sold
- No guarantee of results
- Model performance depends on hardware, model choice and use case
- No remote maintenance as a permanent service
- The goal is knowledge transfer to the client team, not permanent support — “We are done when you are fit.”
AI consulting with a focus on local AI: Typical clients
- Tech leads & development teams
- Questions on architecture, hardware sizing and stack
- Management & C-level
- Compliance status, funding options, entry path
- Regulated / privacy-sensitive organisations
- Need for GDPR- and AI-Act-compliant, locally processed AI
AI consulting with a focus on local AI: Not identical to
- Cloud AI consulting (generic)
- Relies primarily on hosted APIs of large providers; eukalyptusdesign explicitly focuses on on-premise inference
- AI software product / SaaS
- eukalyptusdesign does not sell its own software licence, but consulting and implementation services plus hardware procurement
- Pure hardware sales
- Hardware is procured as part of consulting, configuration and onboarding, not as an isolated product sale
AI consulting with a focus on local AI: Frequently asked questions
What is AI consulting with a focus on local AI at eukalyptusdesign?
A consulting and implementation service of eukalyptusdesign GbR: workshops, prototyping and the procurement and setup of local AI hardware (Apple Silicon, NVIDIA DGX Spark) for companies that want to use generative AI without cloud dependency.
Do clients need their own hardware?
No, not necessarily. eukalyptusdesign can start with existing infrastructure and only recommends hardware when it makes economic and technical sense. If needed, eukalyptusdesign handles consulting, procurement and implementation completely.
How does Apple Silicon compare with NVIDIA DGX Spark?
For local inference with models up to 70B, Apple Silicon (Mac Studio M5 Ultra, 1,200 GB/s memory bandwidth) is often faster than DGX Spark (273 GB/s) because LLM inference is memory-bound. For training above 70B and CUDA migration paths, DGX Spark makes more sense. eukalyptusdesign advises on the right choice for each use case.
Does the training meet the obligation under Art. 4 EU AI Act?
The training is use-case-specific and documentable. It helps companies meet the obligation under Art. 4 EU AI Act, in force since 2 February 2025, to take measures for the AI literacy of their staff (reworded by the AI Omnibus, Regulation (EU) 2026/1744, in force since 27 July 2026; a specific level of competence does not have to be guaranteed) — regardless of whether cloud or local AI is used.
What does a typical proof of concept (PoC) cost?
That depends on the use case, the desired model size and the maturity of the existing infrastructure. A first free tech sparring session (30 minutes) forms the basis for a concrete offer. PoC projects are calculated transparently based on effort.
Is the offer a piece of software or a tool?
No. It is a consulting and implementation service delivered in person. Tools such as MLX, llama.cpp, Ollama or LM Studio are used during delivery — they are aids, not the service offering itself.