Services
Private On-Premise AI
We install and operate generative AI directly in your premises: open-weight models running on GPU servers and AI workstations that we provide on rental, connected to your documents and business systems. Your data never leaves your network, and one accountable team runs the whole stack.
What we take on
Running AI inside your own walls is an infrastructure and integration job, not a subscription — this is the ground we cover.
- 01
Hardware on rental
GPU inference servers and AI workstations provided as a service (Hardware-as-a-Service): sized to your use cases, installed in your premises, maintained, replaced and refreshed by us. No upfront hardware investment.
- 02
Private model deployment
We select, deploy and serve open-weight large language models (LLMs) on your own hardware, behind your firewall. Models are versioned, benchmarked on your tasks and updated on a controlled schedule.
- 03
Assistants on your documents (RAG)
Internal assistants that answer from your own documents through retrieval-augmented generation (RAG), with sources cited and access rights respected document by document.
- 04
Business integration
Connectors to your ERP, CRM and document management (GED/DMS), and AI steps embedded in existing workflows: document extraction, classification, drafting and summarisation.
- 05
On-site technical team
AI and MLOps engineers work in your premises for installation, integration and user training, then return for scheduled on-site visits. Between visits, the platform is operated remotely under SLA.
- 06
Security & access control
Isolated network segment, single sign-on and role-based access, encryption at rest and usage logging. Nobody outside your organisation can query the models or read the prompts.
- 07
GDPR & AI Act documentation
We document data flows, models in use and their purposes, giving you the material you need for your obligations under the GDPR and the EU Artificial Intelligence Act (AI Act).
How an engagement runs
Every on-premise AI engagement follows the same four phases, from the first use case to a platform used every day.
01
Discovery & audit
We map candidate use cases, data sources, volumes, network constraints and the space available in your premises. You receive a written assessment with a prioritised use case and a hardware sizing.
02
Proposal & SLA design
We propose the hardware configuration, the models, the integrations and a pilot with a measurable objective, together with the monthly fee and the P1–P4 service levels for operations. Scope and responsibilities are contractual, not implied.
03
Build / Transition
Our engineers install the servers and workstations on site, deploy the models, connect your data sources and train your users. The pilot is measured against its objective before any roll-out.
04
Run & improve
We operate the platform under SLA: monitoring, model and security updates, hardware maintenance and a monthly usage report. New use cases are added from a shared backlog.
Technologies we work with
We rely on open, proven components that run entirely on your hardware — no dependency on an external AI service.
Models & inference
- vLLM
- Ollama
- Llama
- Mistral
- Qwen
Retrieval & integration
- LangChain
- LlamaIndex
- Open WebUI
- n8n
- Docker
- Kubernetes
Data & hardware
- Qdrant
- pgvector
- NVIDIA CUDA
- Proxmox
- Grafana / Prometheus
- Ubuntu LTS
What you get
The result is an AI your teams use every day, on infrastructure you control, at a predictable monthly cost.
One predictable monthly fee
Hardware, software, integration and operations in a single monthly fee — no upfront investment and no per-request billing that grows with usage.
Data that stays with you
Documents, prompts and answers stay on your network. Confidential files can be used with AI without sending them to an external provider.
Time saved on document work
Searching, summarising, extracting and drafting move from hours to minutes, on the use cases measured during the pilot.
Compliance you can document
Models, data flows, access rights and purposes are documented and kept current, so GDPR and AI Act questions get documented answers.
Frequently asked questions
01Does any of our data leave our premises?
No. The models run on servers installed in your premises, and documents, prompts and answers stay on your network. Our remote operations access is limited to monitoring and maintenance, defined in the contract and logged.
02Do we have to buy the servers?
No. Servers and AI workstations are provided on rental within a monthly fee that also covers maintenance, replacement and refresh. Pricing is quoted after the scoping audit, once the hardware has been sized.
03Which models do you use?
Open-weight models whose licences allow on-premise use, chosen and benchmarked on your own tasks and languages. We have no reseller incentive: the recommendation follows your use case, your hardware budget and your quality requirements.
04Who operates the platform once it is installed?
We do, under our Managed IT Maintenance (SLA) service: monitoring, four contractual severity levels (P1–P4) with defined response times, model and security updates, hardware maintenance and scheduled on-site visits. Hardware, models and operations are covered by one accountable partner.
Let’s talk about your technical operations
A 30-minute introductory call with a senior engineer — no sales script, no obligation. We listen, we ask questions, and we tell you honestly whether we can help.