Deployment
One stack, four postures
Every customer gets an isolated instance with no shared tenancy, and chooses where it runs. The stack is the same in every posture; what changes is where the instance runs, who operates it and, air-gapped, which outside connections exist.
Postures
Where QOR runs
Managed dedicated
We host and operate a single-tenant instance for you.
Best for: Organizations that want the capability without running the infrastructure.
Your cloud
You run the same single-tenant stack in your own cloud account.
Best for: Organizations standardized on their own cloud estate.
On-premises
You run it on your own GPU servers, high-end workstations or compact AI systems.
Best for: Organizations that want the system on hardware they physically control.
Air-gapped
You run it on-premises, installed from an offline bundle with no outside connection.
Best for: Disconnected environments and the most sensitive work.
Air-gapped installations give up outside research, remote nodes, off-site simulation and external model providers, and keep a local counterpart for each. See the details.
Hardware
From a workstation to a cluster
QOR runs on x86_64 and arm64 systems with NVIDIA, AMD, Apple or Intel GPUs. Installations are sized to your workload during scoping. Additional nodes join with their own mutual-TLS identity, on your network or from remote sites over an encrypted fabric, and work is routed to the machine with the right hardware and the most free capacity.
Engagement
What working with us looks like
- 01
Discovery call
30–45 min
A focused conversation about your operational context, data-sensitivity constraints and the workflows where AI could carry real load. No cost, no obligation.
- 02
Scoping engagement
2–3 weeks
A fixed-fee engagement to define target use cases, deployment posture, model selection, hardware or instance sizing, integration requirements and governance boundaries. Covers discovery work, on-site time and travel; fees are credited toward a deployment contract.
- 03
System build & configuration
On-premises: 4–6 weeks, dependent on hardware availability
On-premises, hardware is procured to spec and commissioned; hosted or in your cloud, your dedicated instance is provisioned instead. The platform is configured, and models are deployed and adapted to your terminology and processes.
- 04
Deployment & integration
2–4 weeks
Installation on-site, in your cloud account or in your hosted instance; integration with your internal data sources and systems; operator training; validation against real workflows.
- 05
Ongoing development
Ongoing
Continued refinement as your use of QOR grows: new agents, new integrations, new workflows, and periodic governance and capability reviews.
Every posture
Included in every deployment
Each engagement is scoped in a statement of work. Whatever the posture, it covers:
Co-developed domain specialists
Designed with you against your workflows, vocabulary and decision processes.
Integration engineering
Connectors to your document repositories, laboratory and business systems and instrument APIs, generated from their OpenAPI specifications where they publish one.
Domain model training
Models fine-tuned on your proprietary corpus inside your boundary, once the system is running and a corpus is available.
Ongoing operations
Model updates, new agent capabilities, security patching, tuning and regular capability reviews.