Writing and reasoning
- QwenDense and mixture-of-experts sizes.
- gpt-ossOpenAI’s open-weight models, in 20B and 120B sizes.
- Llama, Mistral & DeepSeekFurther open-weight families.
01 / Built around your work
Your office.
Your models.
Room to grow.
A workstation for one desk. A server for the whole office. A mini PC for a smaller space. We specify, build and install the machine that runs the language models your team needs.
- 1Graphics card chosen for the video memory your models need
- 2Processor document parsing, search and the work around the GPU
- 3Cooler sized for a full working day
- 4Memory room for large models and long documents
- 5NVMe storage fast space for models, search indexes and files
- 6Power supply headroom for sustained full load
- 7Case fans steady airflow under sustained load
- 8Motherboard matched to the parts, with room to upgrade
01 / Workstation
Your everyday
AI co-worker.
Draft, read, search and summarise on one machine. A workstation runs language models beside the tools your team already uses, and keeps your documents on hardware you own.
See what we ran on our own machines
The Office demonstrations ran on one RTX 3090 with 24 GB of graphics memory: Qwen3.8-27B for the writing and reading, Whisper large-v3 for the transcript and nomic-embed-text for the file search, each loaded in turn with every layer on the graphics card. The records keep every prompt, setting and output. Your build is specified so the models you need fit in graphics memory.
- 1Multiple GPUs memory for larger models and more people at once
- 2CPU & memory headroom for your team’s workloads
- 3Fast storage room for models, indexes and shared files
- 4Cooling airflow designed for sustained work
- 5Power sized for the complete GPU configuration
- 6Network connect every desk to shared compute
02 / Multi-GPU server
More GPUs.
Bigger models.
Give the whole office one engine for drafting, search, transcription and automation. Graphics memory largely decides how large a model can run, so a multi-GPU server is where the larger open models live, shared by everyone on your network.
GPU count, model support and memory needs are planned together. We match the chassis, power and cooling to the installation, with room to expand where the platform allows.
Plan my server- 1Processor compact compute for the jobs you need
- 2Memory matched to the software and model size
- 3NVMe storage fast access to your everyday files
- 4Cooling chosen with desk space and noise in mind
- 5Connectivity displays, peripherals and your office network
03 / Mini PC
Small footprint.
Ready for work.
Keep your desk clear and put a compact machine to work. Mini PCs suit drafting, transcription and everyday automation, and with 128 GB of shared memory they load larger models too. Or use one as the desk-side connection to a shared GPU server.
We select the platform around the task. Larger models stay on a dedicated workstation or server while the mini PC gives you a compact place to work.
Plan my mini PCFamiliar models. Served on your network.Qwen · gpt-oss · Llama · Mistral · Gemma · DeepSeek · Whisper
A local model server such as Ollama or llama.cpp runs the models, and your team reaches them from a private interface on your network. These are open-weight model families we can assess for your office. We select and test the models that suit your hardware and the work you want done.
Reading scans and images
- Qwen-VLReads scans, photos and screenshots.
- GemmaGoogle’s open-weight family, with image input.
- Mistral SmallText and image input.
Speech and search
- WhisperSpeech-to-text for recordings and dictation.
- Embedding modelsIndex your files for search, such as nomic-embed-text.
- KokoroText-to-speech voices.
Demonstrated workflows, with model-specific limits. The model, its size, its licence and your hardware decide what a setup can do. Larger models need more graphics memory, and smaller models make more mistakes. We confirm suitability with documents like yours before recommending a setup. This list covers locally runnable model families; cloud model services are separate.
02 / Built for your office
You bring the work.
We build the setup.
Workstations, multi-GPU servers, mini PCs and local AI workflows for the way your office works. A fixed written quote for the complete job.
Try it with Ivan.
Request a hands-on demonstration. Ivan verifies your business and approves a booking.
Agree the setup.
We specify the hardware and itemise the complete job in a fixed written quote.
Build and install.
Brightwork invoices and receives payment upfront before ordering hardware and building your machine.
Hand over.
We test the agreed workflows on your machine and document the setup. Your quote states what training is included.
Four workflows. Installed and tested.
- 01Drafting
- 02Summaries
- 03Search your files
- 04Meeting notes
Supported workflows, tested on your setup with documents like yours. Data extraction, customer replies, translation, coding help and automations are bespoke workflows, scoped and quoted separately.
See how it fits
Your documents in the workflow.
Ivan on the call.
After business verification and booking approval, join Ivan on a video call and try approved workflows on Brightwork’s machines, with documents you’re comfortable sharing.
Request a hands-on demonstrationCustom testing beyond the demonstration can be quoted separately.
Send a demonstration request and we’ll contact you to verify your business and arrange a time. Your booking and temporary access are confirmed separately.
A clear starting point.
| Service | Starting price / rate |
|---|---|
| PC specification, assembly & testingBrightwork specifies and supplies your hardware. | From$654.50+ hardware |
| Office AI installationFour supported workflows, installed and tested on your machine. | From$2,895.75 |
| Custom workflow developmentBespoke work scoped and quoted separately. | From$1,089 |
| Workstation / server upgradesParts itemised for the agreed upgrade. | $214.50/ hour + parts |
| PC maintenanceAvailable only for PCs built by Brightwork. | $214.50/ hour |
Prices above cover the listed service. Hardware, parts, licences and travel are itemised in your written quote where required. PC, server and mini PC builds are quoted for the selected configuration. We agree the complete price and scope before you pay. Hardware orders and builds begin after upfront payment.
Here in person.
Elsewhere, remotely.
In-person installation across Gold Coast, Brisbane and Sunshine Coast. Remote setup elsewhere. Further travel is available when you pay the agreed travel costs.
The handover, clearly scoped.
Your handover covers the installed models and workflows. Maintenance is available only for Brightwork-built PCs.
Your written quote sets out the handover checks, included adjustment rounds and follow-up support period before work begins. Support covers the agreed installed setup. New workflows, added features and later upgrades are quoted separately before work starts.
03 / Who builds it
Talk to the person doing the work.
I’m Ivan. Brightwork brings together the technical setup and the work it needs to serve.
Brightwork takes its name from the polished fittings on a well-kept vessel. Here, it means attention to the parts that make a system useful: setup, clear instructions, careful review and dependable handover.
Your setup starts here.
Try approved workflows with me on a video call. Then get a fixed written quote for the setup your office needs.
About BrightworkYour next step
Try a workflow with Ivan.
Request a hands-on demonstration, or tell us about the setup your office needs.