Skip to main content
gpodz

Guide

What a model risk reviewer will ask about a company-specific model

A model risk or validation reviewer will ask where the training material came from, how versions are tracked, how a change is approved and undone, and how the model is evaluated. At Gpodz, versioning, rollback, separation, and an audit log are designed. Retention and deletion are in progress. Independent attestations are not in place.

Ask Gee about this guide

Gee is an AI demo and can be wrong. Don’t share confidential information. Conversations aren’t stored.

The questions reviewers usually ask

Model risk and validation teams have their own standards, and nothing here replaces them. The questions tend to repeat: Which material shaped the model, and who approved it? Which version is live, and what changed since the last one? Can a change be undone? How was it tested before use? Who can see the record afterward?

Lineage, versioning, rollback, and audit

Statuses below follow the trust page. Designed means part of the platform design and supported by its code, not independently verified.

  • Lineage of training material

    Designed

    Your team selects the material an adapter learns from. The storage layout keeps each organization's material, adapters, and adapter versions under that organization's own paths.

  • Layer versioning

    Designed

    Adapters are versioned separately from the base model, so an adapter can be updated or rolled back without changing the base.

  • Audit log

    Designed

    The audit log is append-only, and each entry is chained to the one before it by hash. Per-organization export is designed but not yet offered.

Approval and evaluation

In the product design, your reviewers test each new adapter version on examples you choose, and it is activated only when you approve it. The trust page does not list these as separate items, and nothing here has been independently verified.

We publish no accuracy or performance figures. Evaluation criteria are yours to set, because you know the work the model must support.

What we cannot show yet

We have no independent validation to hand you. We name no customers without their written permission, so there are no references to call. The role layer is a product design, and Gpodz Compute is pre-launch.

What we can do is answer from the trust page, show you the design, and say plainly where a requirement is not met today.

Data handling and assurance

  • Retention and deletion

    In progress

    Periods and timelines are being set with counsel. Ask us about any deletion request until they are published.

  • Independent security attestations

    Not in place

    We do not hold independent security attestations or certifications today, and no independent penetration test has been completed.

Questions on this topic

Can we roll back a model change?

Rollback is designed. Adapters are versioned separately from the base, so a previous version can be restored without changing the base. It is not independently verified.

Can we export the audit log?

Per-organization export is designed but not yet offered. The log itself is designed to be append-only and hash-chained.

What accuracy figures do you publish?

None. We publish no performance or accuracy figures until we can show how they were measured. Evaluation on your own examples is part of the design.

Is there an independent validation of Gpodz?

No. We hold no independent security attestations and have completed no independent penetration test. If an audit engagement begins, we will list its scope and dates on the trust page.

Bring your review checklist.

We will answer from the trust page and say plainly where we cannot meet a requirement today.

Talk to us