Must have prior experience working with US national security (defense and/or intelligence) agencies-ideally at the level of AI leadership, platform, and compute teams rather than a single program office-and a working knowledge of how national security agencies accredit, budget, govern compute, and measure mission (including the difference between FedRAMP High, IL5/IL6, and classified-fabric accreditation, and the role of cleared cloud providers and system integrators). Help the customer develop evaluation frameworks and measure adoption and value without direct telemetry-provider capacity reports, customer-run dashboards, workload registries, and agreed leading indicators-and guide model migration and release readiness: eval gates built from their workloads, staged cutovers off end-of-support models, and realistic expectations about what lands on their fabric and when.