Verified Engineering is a promise we are willing to be audited on.
Here is the evidence behind it.
Gottlieb is built for engineering organizations where IP protection is a legal and competitive obligation. Every architectural decision, every integration posture, every retention policy reflects that. This page is the evidence, not the pitch.
Engineering principles as moral commitments
These are the constraints the system is built around.
Engineering cannot be bolted on.
Security, lineage, and validation are first-class constraints in the Manufactory, not layers added after the fact. The architecture does not permit an output path that bypasses them.
A render is not a product.
Every Verified Submission carries the reasoning chain, input sources, and measured error bounds that produced it. A deliverable without traceable provenance is not released.
Lineage is a primary artifact.
Who requested it, which data produced it, which models evaluated it, and what a human approved -- all captured at submission time and retained for the full regulatory window.
Verified before quoted.
No cost estimate, compliance assertion, or safety-critical output ships without a completed Verified Submission. The Manufactory enforces this gate; it cannot be bypassed by role or configuration.
Never unattended sign-off in safety-critical settings.
Human approval with a transparent, auditable reasoning chain is mandatory for any output that touches safety-critical parameters. Automation surfaces the evidence; a qualified engineer closes the loop.
Certification and audit evidence
We state what we have. We do not claim what we are still earning.
SOC 2 Type II
Commitment backed: every request is verified.
In progress. Infrastructure controls and access policies are audit-ready. Formal attestation is underway. We share current controls documentation on request.
VPC and on-premises deployment
Commitment backed: your data stays in your region.
Available today. EU customers run on Azure Germany West Central. US customers run on GCP or AWS US-East. On-premises deployment is available for organizations with strict data residency requirements. Production data does not cross regions -- enforced by infrastructure policy, not configuration.
Independent penetration testing
Commitment backed: verified by construction, confirmed by adversarial review.
Planned. The engagement is scoped. Results will be shared under NDA with enterprise evaluators on request.
Data Processing Agreement / BAA
Commitment backed: documented, auditable obligations.
Available on request for any enterprise account.
ITAR and dual-use posture
Gottlieb does not hold an ITAR registration today. Organizations handling ITAR-controlled technical data are responsible for ensuring that data does not enter the Manufactory without appropriate controls in place. We support on-premises deployment and air-gapped configurations for organizations with export-control obligations, and we will engage directly with your compliance team to scope a compliant integration. We will not claim a posture we have not built.
Read-only-first integration
For security-conscious accounts, Gottlieb integrates read-only before any write path is opened. No data is modified, no outputs are committed, and no external systems are called until the integration has been reviewed and the customer has explicitly granted write access. This is the default posture for new enterprise integrations, not an optional mode.
Lineage guarantees under multi-year regulatory retention
Every Verified Submission creates an immutable lineage record: the inputs, the models invoked, the human approver, the timestamp, and the measured error bounds. These records are retained for the full regulatory window required by the customer's jurisdiction -- not deleted on account closure, not truncated at arbitrary intervals. Lineage is traceable, retention-aware, and audit-ready by design.
Liability posture
Gottlieb does not provide unattended sign-off in safety-critical settings. Every output that touches a safety-critical parameter requires a human engineer to review the complete, transparent reasoning chain -- including model confidence, input sources, and measured error rates -- before approval. The Manufactory records that approval and the evidence it rested on. Liability rests with the qualified human who closed the loop, supported by an auditable record of exactly what they approved.