Managed Private AI Systems & Policies.
Inside your boundary.
We deploy the platform into a cloud account you own and are billed for directly. Not a vendor cluster, not a shared tenant. Your data at rest, your records, your exit.
Systems, then policy.
A content policy that cloud identity policy makes mandatory on every request. Roles, quotas and sign-in from day one, and a page inside the product that shows you exactly what is switched on.
Managed, not abandoned.
Releases, checks and upkeep, handled quietly. Direct access to the person who built it when you need it. No support queue, and no lock if you ever leave.
Your private AI, inside your control ·
We walk your architecture, map where your data would flow, and leave you with a written plan for a private AI platform you own. You keep it whether or not you proceed.
A platform that holds its shape.
The product is orgOS: a governed AI environment, deployed as a tagged release into one cloud account that your organization owns. Your people sign in to a dashboard that carries your brand. Every request to a model passes a policy that nobody can switch off from inside an app. Business applications run beside the chat, each on its own isolated origin.
We operate it. You own it. That ordering is the whole design.
Chat and knowledge.
Frontier chat over your own documents, with citations and the limits stated.
Modules.
Business applications inside your environment. Most of them do not need a model at all.
Administration.
Roles, grants and quotas your administrators run, and a short list of things they deliberately cannot change.
The full architecture.
Every layer, what is enforced where, and how the diagram is checked against what was deployed.
Two products. One rule: you own it.
The model works inside your account.
For organizations whose security owner needs AI under their own cloud account's controls. A governed platform with chat, a knowledge base, administration and a module runtime. This is the one the rest of this page describes.
Your files, governed, for the assistant you already use.
For teams already working in a commercial AI assistant. A private server in your own account gives that assistant rule-bound access to your documents and data, computes what must add up, and reports gaps instead of inventing. A different data boundary, stated plainly.
Not sure which? Platform or gateway: an honest way to choose.
Residency where it matters. Frontier everywhere else.
Canadian data residency is what we deploy for - and it is not all-or-nothing. Pin the workloads that need pinning, take the frontier models everywhere else, in the same deployment. Every model below says where its inference actually runs, so that call gets made per workload instead of being imposed once on the whole platform. The catalogue is the full one, at wholesale, across both of the managed model service's inference planes.
Anthropic
13- Claude 3 Haiku
- Claude 3 Sonnet
- Claude Fable 5
- Claude Fable 5.1
- Claude Haiku 4.5
- Claude Opus 4.5
- Claude Opus 4.6
- Claude Opus 4.7
- Claude Opus 4.8
- Claude Opus 5
- Claude Sonnet 4.5
- Claude Sonnet 4.6
- Claude Sonnet 5
OpenAI
7- GPT OSS Safeguard 20B
- GPT OSS Safeguard 120B
- GPT-5.6 Luna
- GPT-5.6 Sol
- GPT-5.6 Terra
- gpt-oss-20b
- gpt-oss-120b
xAI
1- Grok 4.6
Meta
7- Llama 3 8B Instruct
- Llama 3 70B Instruct
- Llama 3.1 8B Instruct
- Llama 3.1 70B Instruct
- Llama 3.3 70B Instruct
- Llama 4 Maverick 17B Instruct
- Llama 4 Scout 17B Instruct
Mistral AI
14- Devstral 2 123B
- Magistral Small 2509
- Ministral 3 8B
- Ministral 3B
- Ministral 14B 3.0
- Mistral 7B Instruct
- Mistral Large (24.02)
- Mistral Large (24.07)
- Mistral Large 3
- Mistral Small (24.02)
- Mixtral 8x7B Instruct
- Pixtral Large (25.02)
- Voxtral Mini 3B 2507
- Voxtral Small 24B 2507
DeepSeek
3- DeepSeek V3.2
- DeepSeek-R1
- DeepSeek-V3.1
Qwen
7- Qwen3 32B (dense)
- Qwen3 235B A22B 2507
- Qwen3 Coder 480B A35B Instruct
- Qwen3 Coder Next
- Qwen3 Next 80B A3B
- Qwen3 VL 235B A22B
- Qwen3-Coder-30B-A3B-Instruct
Moonshot AI
2- Kimi K2 Thinking
- Kimi K2.5
MiniMax
3- MiniMax M2
- MiniMax M2.1
- MiniMax M2.5
Z.AI
3- GLM 4.7
- GLM 4.7 Flash
- GLM 5
NVIDIA
4- Nemotron Nano 3 30B
- NVIDIA Nemotron 3 Super 120B A12B
- NVIDIA Nemotron Nano 9B v2
- NVIDIA Nemotron Nano 12B v2 VL BF16
- Gemma 3 4B IT
- Gemma 3 12B IT
- Gemma 3 27B PT
Writer
3- Palmyra X4
- Palmyra X5
- Writer Palmyra Vision 7B
Amazon
11- Amazon Nova Multimodal Embeddings
- Nova 2 Lite
- Nova 2 Sonic
- Nova Lite
- Nova Micro
- Nova Pro
- Rerank 1.0
- Titan Embeddings G1 - Text
- Titan Embeddings G2 - Text
- Titan Multimodal Embeddings G1
- Titan Text Embeddings v2
Cohere
4- Embed English
- Embed Multilingual
- Embed v4
- Rerank 3.5
TwelveLabs
2- Marengo Embed 3.0
- Pegasus v1.2
Stability AI
16- Stable Diffusion 3.5 Large
- Stable Image Conservative Upscale
- Stable Image Control Sketch
- Stable Image Control Structure
- Stable Image Core 1.0
- Stable Image Creative Upscale
- Stable Image Erase Object
- Stable Image Fast Upscale
- Stable Image Inpaint
- Stable Image Outpaint
- Stable Image Remove Background
- Stable Image Search and Recolor
- Stable Image Search and Replace
- Stable Image Style Guide
- Stable Image Style Transfer
- Stable Image Ultra 1.0
Luma AI
1- Ray v2
Nothing in the catalogue matches that filter.
Verified against the real thing.
After a deploy, we attack our own work before you rely on it.
Unit tests prove the code agrees with itself. They do not prove a live deployment refuses a forged session. So after a release lands in your account, acceptance scripts exercise the real thing over HTTPS and every one of these must be refused.
- 01Refused by policy, not by convention. A request without your content policy is denied by cloud identity policy before it reaches a model. Tests fail our build if that deny is ever weakened.
- 02What you read is what is built. The page inside the product that lists your controls, and your standing exceptions, is generated from the same object that builds the infrastructure.
- 03Nothing reaches production except by hash. A named administrator approves the exact artifact. Our own automation cannot promote.
You can see how it is set up. That is exactly what lets you answer for it.
Most private AI means they hold your data privately. Ours means you do. Everything else on this site follows from that one sentence.
- 01Architecture. Everything is code in your account, and your diagram is checked against the infrastructure that was actually deployed.
- 02Cost. Your cloud provider bills you directly for compute. We never sit between you and that bill, so there is nothing for us to mark up.
- 03Relationship. The monthly fee buys the service, not access. If we part ways, you log into your own cloud and keep running.
Priced like infrastructure, not like seats.
The bounded first step. A written plan you keep whether or not you proceed.
orgOS stood up in your own account. Typically lands around $5,000 once the work is scoped.
Releases, checks, upkeep and a direct line. Cancel anytime. You lose us, not the platform.
Modules, automation and forward-deployed engineering, each against acceptance criteria written first.
Compute is billed by your cloud provider directly to you at its list price. Full pricing · What compute actually costs · The engagement path, gate by gate
One argument, different doors.
The security owner carries the risk if they say yes. The operating owner carries the cost of saying no. Meanwhile the work goes on anyway, in browser tabs, with no governance at all. That standoff is what we resolve, and each person in it needs a different first answer.
Security and privacy.
Not opposed to AI. Opposed to uncontrolled AI. More control, not less.
Operations.
Permission turned into work that actually gets done, measured by outcomes you choose.
Executives and owners.
Your business does not need another AI tool. It needs an operating capability it owns.
Finance.
Bounded exposure with stage gates. A compute bill you can always see.
Running it yourself? The page for IT leaders is the most technical of the five.
A small senior practice with a factory.
org.tech is not a large company, and it does not pretend to be. It is a founder-led practice whose reach comes from a versioned product and a deployment factory rather than from headcount. The person you talk to is the person who built it.
That only works if the engineering discipline is real. So the numbers on the right are about the product, not about us. About org.tech · What happens if we are not here
Field notes from the boundary.
Why your models run inside your cloud, not ours.
Most "private AI" still means a vendor holds your data privately. Whose account the platform runs in is the one design choice that changes every downstream decision.
The security owner carries the risk of yes. The operating owner carries the cost of no.
Why a ban never produces zero AI, and what actually resolves the standoff.
Data residency is a choice with a price. Here is the price.
Storage residency, inference residency, and what pinning to Canada costs you in model choice.