Sentinel Protocol LLC
Making Autonomous AI Accountable

Rami Khaldy, Founder

Sentinel Protocol LLC was built around that problem.

It does not develop another foundation model. Instead, Sentinel builds the operating environment around frontier intelligence, connecting it with mission state, evidence, specialist responsibilities, verification, authority, handoffs and controls.

“Models supply intelligence. Sentinel organizes its use. They do not share a mind. They share a record,” says Rami Khaldy, Founder of Sentinel Protocol LLC.

Engineering Behind the Approach

The company’s thinking comes from Khaldy’s career in instrumentation, test automation and systems integration, together with his service in the U.S. Army Reserve. His engineering work has included CF-18 avionics, National Instruments, Rolls-Royce engine research, F-35 systems, C4I integration, NASA Artemis test-conductor operations and flight-test instrumentation.

Across those settings, the questions were practical: Did the instrument actually measure what it was supposed to measure? Can another engineer reproduce the result? What evidence supports the conclusion? What does the next person need to continue?

Those questions eventually followed Khaldy into his work with increasingly capable AI.

At first, the engineer was effectively the control system— maintaining context, checking evidence, coordinating specialized intelligence and deciding when work could proceed. Sentinel emerged when Khaldy recognized that this synchronization function had to become architecture without reducing the capability of the intelligence itself.

The result evolved into a persistent artificial engineering organization in which models can change while mission state, evidence, authority, verification and organizational learning endure.

Sentinel’s goal is not to make AI sound more certain. It is to keep evidence, responsibility and continuity attached to the work.

Sentinel works across frontier providers rather than depending on a single model. Where a provider’s native coding, tool-use or agent environment is strong, Sentinel can work within it while connecting that intelligence to a persistent mission layer.

That distinction is important: the model supplies intelligence; it does not automatically become the mission authority.

A frontier provider controls its models and platform. The organization using them retains control of its mission, protected information, authority structure, verification requirements and consequential decisions. Sentinel creates the boundary between the two.

This also supports protected environments. Frontier models can help design logic, interfaces, tests and software hooks using synthetic or sanitized information. Approved local agents can then use those interfaces inside the authorized environment with actual protected data.

The principle is simple: the design can travel; the protected data does not.

Sentinel also places governance at the action layer. Tool calls and consequential changes can be checked against permissions and policies before they proceed. Specialist roles separate implementation from independent challenge and verification.

A check must be capable of failing to demonstrate that it can detect the condition it claims to detect.

Keeping the Mission Moving

The discipline becomes especially clear in the Operations Conductor. Khaldy first articulated the automated test-conductor concept during his NASA assignment, where unexpected events could scatter important information across instruments, procedures, engineering systems, communications and reports.

Sentinel’s approach brings those pieces into a working view while keeping source and time attached.

If two instruments disagree, the system does not manufacture agreement. Both observations remain visible while the appropriate specialists investigate. Human authority remains responsible for the consequential decision.

“The mission record is the unit of continuity. A transcript is not,” Khaldy says.

The principle also shapes handovers. When responsibility changes—or even when the underlying model changes—the next contributor should inherit the current condition, supporting evidence, decisions already made, unresolved obligations and a route to resume. The organization should not have to reconstruct the mission from a conversation.

That produces another important characteristic of Sentinel: the intelligence can change without requiring the organization to forget how it works.

A Practical Path for Defense AI

Sentinel is currently working with a government partner on a test-conductor-inspired AI capability that is moving from concept toward a Sentinel Protocol product, applying the same mission-assurance principles to operational environments where evidence, continuity and human authority are critical.

Sentinel’s architecture is designed for environments where powerful AI must coexist with evidence, verification, security, continuity and accountable human authority.

The same foundation can support aerospace testing, engineering control rooms, transportation, critical infrastructure and other mission-critical institutional workflows.

The larger opportunity is practical: allow organizations to use increasingly capable frontier intelligence without surrendering command of the mission to the intelligence or its provider.

Sentinel Protocol LLC has been named Top Aerospace and Defense AI Governance Solution 2026, recognizing an approach rooted in the engineering discipline of keeping intelligence connected to the evidence, mission and people responsible for it.