Senior AI Policy Engineer: Secure Out-of-Band Proxy

Redpanda Data Inc.; Polska; Hybrid; PLN 802,000 - 899,000; Full time

Job summary

Redpanda Data Inc. is seeking a security-minded engineer to move the out-of-band policy engine into production.

You will wire a Cedar-based enforcement proxy into the AI gateway to govern all tool calls, with two-tier policy composition and per-agent narrowing to ensure secure, auditable decisions. You will extend policy operators, support deferred-approval workflows, and collaborate with global customers to shape the product.

Qualifications

• 5+ years of experience building systems that enforce authorization in live request paths (proxy/gateway/middleware).
• Experience with policy-as-code systems such as Cedar, Open Policy Agent (Rego), XACML, or comparable frameworks.
• Solid understanding of AI agent and LLM failure modes and defenses.
• Comfortable with adversarial testing mindset and production environments.
• Experience collaborating with globally distributed teams via GitHub and in the open.
• Strong verbal and written communication and technical ownership.

Responsibilities

• Take the policy enforcement proxy from prototype to production within the AI gateway.
• Develop a two-tier policy model combining per-tool-server ceilings with per-agent narrowing.
• Extend policy operators across authorization, data exposure, and semantic gating including deferred-approval workflows.
• Ensure enforcement decisions are auditable and produce traceable evidence for customers.
• Maintain and extend adversarial certification harness with red-team scenarios and measurable leakage metrics.
• Work with security-conscious customers to translate needs into product capabilities.
• Raise systemic challenges to management and influence the roadmap.

Skills

Adversarial testing

Security policy thinking

Distributed teamwork

Technical ownership

Tools

Cedar

Open Policy Agent (Rego)

XACML

Job description

Redpanda Data Inc. is seeking a security-minded engineer to move the out-of-band policy engine into production.

You will wire a Cedar-based enforcement proxy into the AI gateway to govern all tool calls, with two-tier policy composition and per-agent narrowing to ensure secure, auditable decisions. You will extend policy operators, support deferred-approval workflows, and collaborate with global customers to shape the product.