Run document AI in your own environment
A bounded private document assistant with measured performance and a reproducible deployment.
Pricing preview · proposed fees and scope await approval.
What we deliver
- One open-weight model, one document task and one client-controlled environment
- A 30-case benchmark reporting task quality, latency and resource use
- Access controls, network checks, rollback and backup restoration
- Deployment code, dependency and license inventory, runbook and team training
How we check the work
- Recreate the deployment from its pinned configuration.
- Run the agreed task set with outbound access blocked after provisioning.
- Record performance on your hardware; demonstrate stop, rollback and restore.
What you bring
Suitable provisioned compute, infrastructure access and approved model licensing. Feasibility and capacity are checked before booking.
What the fee covers
The fixed fee covers the scope above. A fit check confirms eligibility, not a different price. We agree on any separate work before it begins.
Hardware, cloud bills, model training, 24/7 operation and compliance certification are outside this scope.
What working together looks like
Bring one real task
Use the material and tools you already have. We agree on the output, record a baseline and decide what a good result must pass.
Build it together
We build the scoped workflow, run representative cases and compare the results against your reference work. Exceptions become explicit checks.
Let your team take over
Someone on your team runs a new case, explains the checks and teaches a colleague. You leave with editable work, operating instructions and a named owner.
The handover covers how to run, inspect, stop and maintain the delivered workflow. Ongoing support is outside the fixed scope.
Keep the work. Know how to run it.
You receive the project’s custom code, editable files and team handover. Open-source dependencies retain their licenses. We use your existing tools where suitable; any model usage, cloud or data-license costs are identified before work begins. There is no required KCQ software subscription in these proposed packages.
Explore a first output
These online examples prepare drafts and supported calculations. Connecting and deploying the workflow is part of the engagement.
Run an open-weight model in our cloud for confidential documentsWhy this work · published market examples
These sources document the workflow category. They are not KCQ customers, endorsements, or guarantees of equivalent results.
Salt · Platform documentation · Product offering
Salt describes connecting enterprise data to multi-step AI workflows deployed in customer-controlled environments.
Interdimensional · Training programs · Service offering
Training is paired with implementation, explicit acceptance checks and operational handover.