Contact
Start with the diagnostic.
$7,500 fixed · ten business days · one workflow
Credited in full against a build signed within 60 days.
Engagement
How an engagement is priced.
Diagnostic
$7,500
Ten business days, one workflow. Credited in full against a build signed within 60 days.
Production build
From $25,000
Fixed price, accepted against tests written before the build.
Operate
From $2,500 a month
Bounded and optional.
How it starts
One workflow, with its data boundary written down before anything is built.
Send a short brief
The workflow, where it breaks today, and what a wrong or slow answer costs you.
We assess fit
We reply within one business day. If the workflow is not a fit, we say so.
20-minute call
If it fits, we book a 20-minute call to scope the diagnostic.
Need a website, SEO or ads rather than an AI system? Contact OptiVis Marketing.
- info@optivis.dev
- Phone
- 281-836-4511
- Office
- 13518 West Bellfort Ave, Suite 200
Sugar Land, TX 77498
Response within 1 business day.
Commerce operators and legal and professional firms. Based in Sugar Land, Texas.
Written data boundaries
The boundary is a document, one per engagement.
“Private” is a claim a procurement reviewer should test, so we write it down. For every system we build, a data-boundary document states where each component runs, which model calls leave your environment and to which provider, and how OCR, embeddings, logs, backups and support access are handled, with retention and deletion. You get the first version in the diagnostic, before anything is built.
Client data is processed only within the deployment boundaries agreed in writing. Any external provider, data transfer, and retention policy is documented before use.
- Component placement
- Where each component runs
- Model calls
- Which provider receives which data
- OCR
- Where processing occurs
- Embeddings
- Where vectors are stored
- Logs
- What is retained
- Backups
- Where copies reside
- Support access
- Who can access what
- Retention and deletion
- Duration and removal procedure
These are the fields each document answers, not a record of a particular deployment.
One retailer
Commerce is an application of the work, not a practice.
An auto-parts retailer whose Shopify store we build and run. One measured window, both denominators, and the limit of what it shows.
Shopify session report, Q3 2026: ChatGPT, 19 orders from 151 sessions, 12.58%; Google search, 76 orders from 3,988 sessions, 1.91%.
The limit. Attribution shows where orders came from. It does not prove profit, and it does not prove that our work caused them. Small sample, one store, one quarter.
- Status
- We build and run the store.
- Boundary
- The client’s store and the client’s Shopify data.
- Acceptance
- Re-measured from Shopify attribution each quarter, whatever it shows.
The evaluator
We taught a smaller model to make the same calls as the evaluator.
The evaluator is Jev, from TypeSafe AI. The smaller model is ModernBERT-large, distilled from 68,828 Jev-labeled training examples, so the same calls run on compute we operate.
It is measured as teacher agreement, not accuracy, on two cohorts, with zero-shot Laya as the baseline.
In-KB n = 4,387
Unseen KB n = 1,301
The limit. Rare flags are not claimed.
- Status
- Measured. The figures are published on OptiVis AI Labs.
- Boundary
- Trained and run on GPU compute we operate.
- Acceptance
- Teacher agreement on two cohorts, against a zero-shot baseline.
Engineering standard
One standard under every system.
Evaluation
A system is measured against a baseline before we call it done, the way the distillation above was measured against zero-shot Laya.
Evidence-backed retrieval
Answers point to the passages they came from. A quote that does not match its source is refused, not shown.
Written acceptance
Acceptance tests are written down before the build, and the build is accepted against them.