Description
The Agent Development Platform is a production-grade, cloud-native ecosystem that enables engineering teams to define, orchestrate, deploy, and observe AI agents at scale. It standardizes agent development across the organization using LangGraph and Strands Agents on AWS AgentCore Runtime, with authorization, security, quality, and governance controls for interactions between agents.
Responsibilities
•Lead the architecture of the policy engine and Cedar policy schema across the platform
•Define authorization models for agent-to-agent capability advertisement and invocation
•Drive enterprise InfoSec review processes and align them with ARB requirements
•Establish default-deny authorization patterns for agent interactions
•Integrate identity-aware authorization using OAuth 2.0, JWT, and MS Entra
•Build and maintain policy-as-code patterns to standardize governance across agent development teams
•Oversee audit logging for policy decisions and enforcement outcomes
•Guide staged policy enforcement rollouts, from monitoring to full enforcement
•Collaborate with engineering teams to reason about trust boundaries between interacting agents
Requirements
•5+ years of experience in cloud security or identity engineering
•Hands-on experience with authorization or policy-as-code for AI agents in a production agentic AI project, such as Cedar/OPA policy engines enforcing agent-to-agent boundaries, capability/agent-card discovery, or default-deny agent authorization models
•Expertise in identity-aware authorization, including OAuth 2.0, JWT token inspection and claims mapping, and OIDC
•Skills in RBAC and ABAC design patterns
•Experience with policy-as-code (Cedar, OPA) or demonstrable A2A protocol depth, including agent identity, capability/agent-card discovery, and trust boundaries between interacting agents
•Experience leading enterprise security review processes, including InfoSec and ARB workflows
•Knowledge of AWS AgentCore Policy and Cedar policy language (principals, actions, resources, conditions)
•Proficiency in English at a B2+ level
Nice to have
•Familiarity with Cedar or AWS Cedar (open source)
•Practical familiarity with policy-decision-point patterns including request evaluation against declarative rules, default-deny models and staged rollout from LOG_ONLY to ENFORCE
•Early experience with AWS Verified Permissions or AgentCore Policy
•Cloud-native AWS exposure
•Skills in zero-trust architecture design
Conditions
No conditions specified in the posting.
