Verified EngineeringBook a call

Trust and security

AI data programs carry your models, your prompts and your roadmap. Here is exactly how an engagement runs, and what protects your work at each step.

How an engagement runs

  1. NDA before anything else

    We sign your NDA, or ours, before you share a single file, prompt or brief.

    What protects youNothing changes hands until it is signed.

  2. Scope and acceptance criteria

    A short call on the data type, schema, standard (ASME or ISO), volume and what “correct” means for you. We write it down as acceptance criteria.

    What protects youOnly the people named on your program join the call and see the brief.

  3. Fixed quote and contract

    A price per model or per task, a delivery date for the pilot batch, and the IP terms.

    What protects youWork product is assigned to you as the contract specifies; we agree up front which internal tools stay ours.

  4. Secure exchange

    Reference files and briefs move through a project folder created for your program.

    What protects youAccess is limited to the engineers on your program and removed when it ends.

  5. Pilot batch

    We produce a small batch through the full pipeline, with a validation report for every item.

    What protects youEvery item is checked by scripts and by a second engineer before it reaches you.

  6. Your review

    You test the batch in your own pipeline. Anything that misses the acceptance criteria is fixed at our cost.

    What protects youYou decide whether to continue; there is no commitment beyond the pilot.

  7. Scale up

    Agree monthly volume, parallel tracks and a ramp-up plan. We add engineers to meet it.

    What protects youThe same controls apply at every volume, and we never name you or describe your program without written permission.

Pricing and pilot details are on the pilot page.

Standing safeguards

  • Certified information security

    Our operations are certified to ISO/IEC 27001:2022 and ISO 9001:2015 by QRO (EGAC-accredited), valid to January 2028. Certificate copies and scope on request.

  • Verified human experts

    Every model is made and checked by qualified engineers on our own staff. No crowd work and no unreviewed AI output in your data.

  • Confidentiality by default

    We never name clients or describe their programs without written permission.

Need our security questionnaire answered?

Send it over. We’ll return it with our certificates and the relevant policies.