Small, focused, by design.

Althorn is a boutique consultancy focused on private, self-hosted AI. Specialist work for businesses that take data seriously.

How this started

Althorn grew out of years of self-hosting. The open-source stack we install for clients is the same one we've been running at home for years: AI assistants, voice control, private search, photo libraries, encrypted backups. Over thirty services in total. We use what we sell.

Big consulting firms charge $20K and up for private AI. Cloud AI is cheap but your data leaves the building. Althorn sits in the middle: specialist work at fair, transparent prices, for businesses that need this but don't need a vendor to walk them through a six-week discovery phase.

Fewer clients, deeper work. That's the whole pitch.

Your data stays put

Nothing leaves your office. Not for training, not for analytics, not for anything. Full stop.

Open source tools

Everything we use is open source. No vendor lock-in. If you wanted to take over and run it yourself one day, you could.

No jargon

We explain everything in plain English. And if AI doesn't make sense for your business, we'll tell you that too.

Brisbane-based

Local. Not a call centre. Not overseas. We're in Brisbane and we understand how Australian businesses work.

What you actually get

No jargon. Here's what every setup looks like in practice.

Runs on your hardware

The AI runs on a machine in your office, not on someone else's cloud. Nothing leaves the building unless you want it to.

Best-fit open-source AI

We deploy Gemma, DeepSeek, GPT-OSS, Qwen, Llama, and other open-weight models. When better ones come out, we swap them in. No licensing lock-in.

Familiar chat interface

Feels like ChatGPT. Your team picks it up in minutes, no training manual required.

Fast for the whole team

Hardware-accelerated so several people can use it at once without waiting on each other.

Encrypted, on-premise backups

Automatic backups, encrypted, stored locally. If hardware fails, nothing is lost.

Hardened and maintained

Full-disk encryption, regular security patches, and monitoring so the system stays healthy over time.

The technology we deploy

Open-source, battle-tested, and trusted by organisations worldwide. No vendor lock-in, no per-seat licensing.

AI Models

Gemma, DeepSeek, GPT-OSS, Qwen, Llama, and Mistral. We match the right model to your workload and hardware. When a better model ships, we swap it in. No licensing lock-in.

Inference

Ollama and vLLM for model serving. Hardware-accelerated on NVIDIA GPUs with CUDA. Your team gets ChatGPT-level speed without the cloud.

Interface

Open WebUI. A clean, familiar chat interface that works in any browser. Feels like ChatGPT. Your staff need zero training.

Hardware

NVIDIA GPUs, from single-card workstations to multi-GPU servers. We spec, source, and build the right hardware for your team size and workload.

Security

Full-disk encryption (LUKS), network segmentation, role-based access, and audit logging. Every deployment is hardened before it goes live.

Document Search

Retrieval-augmented generation (RAG) connects the AI to your internal documents. Ask questions about your own files. Answers stay on your network.

We run and deploy the same open-source tools ourselves. 31 services, zero cloud dependencies.

NVIDIA Ollama vLLM Open WebUI Home Assistant Nextcloud Immich Navidrome AdGuard Home SearXNG Uptime Kuma Firecrawl Music Assistant Comet PostgreSQL MariaDB Redis RabbitMQ

Want to see it running?

We can show you a live demo on a 20-minute call. Same setup we'd put in your office.

Book a Consultation