A change can be authorized and still take production down.
Permissions answer whether an actor may change production. Nothing answers whether this change should happen in this environment right now — or governs the authority to carry it out, or checks afterwards that reality matched the decision. ChangeGuard AI is that layer: the decision and control layer for infrastructure change.
Infrastructure is changed by people, pipelines and now AI agents.
Change volume scales with machines. Judgment does not. Every one of those actors gets the same treatment: a judgment before the change, authority bound to a Permit, an Executor that carries out only the permitted action, and an outcome verified against reality.
Judgment → Permit → Execution → Verified outcome
What we’re building
A system that judges a proposed change against the actual target environment, binds authorized execution to a short-lived Permit, lets an Executor carry out only the permitted action, and verifies what actually happened. Judgment, Permit, Executor, Observation — four things, kept apart on purpose.
Where it stands
Live-proven end to end. The Permit, the Cloud Executor, environment-scoped execution identity, AWS Account Center and Connected Environments run in production, exercised end to end against the live system and guarded by blocking security tests. We are now opening production evaluations with a small number of design partners.
How we build it
Authority over production is earned, never assumed. Every capability ships behind evidence and human control; every claim is labeled live, live-proven, supported, roadmap or vision.
“I spent fifteen years being the engineer who made these judgments by hand.”
Kristopher White, Founder & CEO. Fifteen years inside production infrastructure — infrastructure, DevOps, cloud and technology risk at SAP NS2, the U.S. Census Bureau, Johns Hopkins University, NASA Goddard and Capital One.
- Led the platforms this product judges: Lead Cloud Solutions Architect at the U.S. Census Bureau — OpenShift, EKS and Helm deployments, GitHub Actions automation, infrastructure risk management
- Operated change at scale: continuous deployment for 1,000+ containers behind Johns Hopkins University and Hospital applications, plus AWS migration work
The workflow ChangeGuard AI replaces is the one he ran by hand — which is why it knows exactly where it hurts. Talk to the founder →
Principles we build against
- Evidence over assertion. Every claim ties to a real signal; uncited reasoning is dropped, not shown.
- Authority is governed, not assumed. Nothing executes without a Permit; nothing is VERIFIED without observation.
- Earn the critical path. Advise by default; Auto only inside a policy you wrote.
- Label the truth. What is live, what is proven, what is roadmap and what is vision are never blurred.
ChangeGuard AI is a member of the NVIDIA Inception program.
Kubernetes on AWS EKS first. Infrastructure change is the object.
Kubernetes on EKS is where change is most frequent and most consequential, and it is live-proven today. The same judgment, Permit and verification extend to every surface a production change can originate from — earned domain by domain on evidence, and labeled as roadmap until it is real.
Come see it decide.
Watch it judge a real change — then point it at one of your own environments.