Principal Security Architect, Agent Policy Fabric

NVIDIA•Santa Clara, TN

1d•$272,000 - $431,250

About The Position

NVIDIA is seeking a Principal Security Architect, Cloud Engineering & Services, to join our Agent Security, Safety, and Governance team. This role will lead cross-company security architecture for agentic AI, using Agent Policy Fabric as a starting point for enterprise agent governance. The architect will coordinate efforts among various teams including security product, OpenShell, runtime groups, Identity, IT, Fleet/MDM, SecOps, corporate application owners, and partner organizations to develop practical security standards, roadmaps, and adoption plans from a working draft architecture.

Requirements

Bachelor's degree (or equivalent experience) with 15+ years of industry experience in security architecture, product security, enterprise security platforms, identity and access management, cloud security, or infrastructure governance.
Validated ability to lead ambiguous, cross-functional security initiatives across product, platform, infrastructure, IT, and security operations teams.
Practical understanding of agentic AI risks, tool-call governance, prompt-injection limits, sandbox boundaries, credential exposure risks, audit requirements, and the difference between containment, authorization, and monitoring.
Experience designing controls around identity, authorization, policy, secrets, network egress, runtime isolation, telemetry, SIEM integration, exception workflows, and compliance evidence.
Ability to write crisp architecture memos, decision records, threat models, standards, and adoption plans that are useful to both senior leaders and implementation teams.

Nice To Haves

Experience securing agent platforms, AI copilots, autonomous workflows, MCP-style tool systems, sandboxed runtimes, or governed access to enterprise SaaS and engineering systems.
Familiarity with OPA/Rego, Cedar, Zanzibar-style authorization, OAuth/OIDC, SAML, workload identity, delegated authorization, signed configuration, or enterprise trust-root distribution.
Track record driving company-wide security architecture across multiple business units, including standards, rollout plans, risk acceptance, exception handling, and measurable adoption.
Experience explaining security architecture to executives, customers, partners, standards bodies, or field teams while preserving bounded claims and clear implementation caveats.

Responsibilities

Lead Enterprise Agent Security Architecture: Define the cross-company reference architecture for governed agent actions, including durable policies, runtime controls, adapter boundaries, credential mediation, detector response, audit correlation, failure modes, and production-readiness criteria.
Drive APF as a Governance Starting Point: Translate Agent Policy Fabric concepts into executive-ready decision papers, engineering standards, threat models, control objectives, and implementation achievements without treating working-draft architecture as a pre-decided product direction.
Align Cross-Organization Owners: Partner with Product Security, OpenShell, Omnistation, Identity, IT, Fleet/MDM, SecOps, 3S, legal/privacy, and corporate-resource owners to define who owns each control surface and how agent workflows move from proof-of-life to enterprise pilot.
Build Security Review and Adoption: Establish review patterns for agent workflows, including policy authoring, approval, signing, runtime admission, credential issuance, direct-egress controls, audit evidence, managing anomalies, and break-glass procedures.
Represent the Architecture: Brief senior leaders, customer-facing teams, and partner engineering teams on NVIDIA's agent security posture, APF maturation path, open decisions, known limitations, and the evidence required before broader deployment.