Netics
All services
Private AI on Your Own Hardware

AI that runs in your office, on hardware you own.

We size, install and maintain the machine that runs your AI on site: a compact Mac mini for a small team, an AMD or NVIDIA machine for heavier work, up to a Mac Studio for the largest open models. Your documents stay in the building, and Netics Harness and Scribe run on top.

Send a message

What's included

Hardware Sizing & Consultation

We match the machine to your models, users and budget, with the honest trade-offs: memory, speed, noise, power and price.

Local Model Setup

Open models such as Qwen installed, quantized and tested on your hardware, then connected to Harness, Scribe or your own tools.

Maintenance Contract

Updates, model upgrades, monitoring and backups for your AI machine, for a fixed monthly fee.

Diagram of private AI on your own hardware: inside your office, your team uses Netics Harness and Scribe on a local AI machine running an open model over your own documents; a routing rule keeps anything confidential local and may send only non-sensitive tasks to an optional API provider; Netics maintains the machine remotely.
Scroll sideways to see the full diagram.

Hardware we set up

We size each machine to your models and users. Hardware and models change fast, so we recommend a configuration at the start of each project.

Mac mini

Memory
Entry configurations
What it runs
Small and mid-size open models
Best for
A quiet office assistant for a small team

Mac mini with a Pro chip

Memory
More unified memory
What it runs
Mid-size open models such as Qwen, at a comfortable speed
Best for
A private document assistant for a team

AMD Ryzen AI mini PC

Memory
Large unified memory
What it runs
Mid-size models fast, very large models at a slower pace
Best for
A lot of memory for the budget, Linux friendly

NVIDIA DGX Spark

Memory
Large unified memory
What it runs
Large open models with the full NVIDIA software stack
Best for
Teams that also fine-tune their own models

NVIDIA RTX workstation

Memory
Dedicated GPU memory
What it runs
The fastest answers on mid-size models
Best for
Many users at the same time

Mac Studio

Memory
Very large unified memory
What it runs
The largest open models, entirely on site
Best for
Everything stays in the building, at full scale

Works with

  • Apple
  • NVIDIA
  • AMD
  • Ollama
  • Hugging Face
  • Qwen (Alibaba Cloud)
  • Mistral AI
  • Meta Llama
  • DeepSeek

A good fit if

  • Your documents must stay on your premises
  • You prefer a fixed hardware cost to a growing API bill
  • Your AI has to keep working offline

Related solution

Netics HarnessYour own AI agent platform. Any model, your rules.

Talk it through with an engineer

Thirty minutes to look at your situation and what would help. You leave with a clear next step, whatever you decide.

Send a message