Engineer, AI Prompt Security

Job Title

GCS – Security Architecture Function AI Engineer

The GCS – Security Architecture function secures Carnival Corp's adoption of AI by protecting GenAI and agentic systems at the prompt and tool-invocation layer. This position is a hands-on engineering role centered on scripting, programming, and automation—building the guardrails, detections, evaluation tooling, and reusable libraries that defend large language model and agentic applications against prompt injection, jailbreaks, and tool misuse. The AI Engineer writes the software and automation that ships these protections into CI/CD and engineering workflows and operationalizes findings from the automated AI red-team / security posture platform to reduce risk to Carnival Corp and customer information assets.

• Design, build, and maintain prompt-level security controls and guardrails (LLM firewalls, input/output filtering, jailbreak and prompt-injection defenses) for GenAI and agentic applications.
• Develop scripts, programs, and automation (primarily Python) to operationalize prompt-security controls, detections, and policy enforcement across AI applications.
• Build and maintain automated evaluation harnesses and regression test suites that continuously validate prompt and agent behavior against security policies.
• Integrate prompt-security controls and tests into CI/CD pipelines and engineering workflows so protections ship with every release.
• Build reusable libraries, SDKs, and templates that make secure prompting and safe tool-use the default for engineering teams.
• Enforce identity and PHI/PII/data-protection controls at the prompt and tool-invocation layer, ensuring agents respect least-privilege and data-handling boundaries.
• Define and document secure-prompting and agentic-development best practices, patterns, and anti-patterns, and drive adoption across engineering teams.
• Operationalize findings from the automated AI red-team / security posture platform, translating them into prompt-layer fixes and detections.
• Research and pilot prompt-security and LLM-defense tooling aligned to the NIST AI RMF and OWASP Top 10 for LLM.

• Strong software-engineering fundamentals—scripting, programming, and automation (especially Python)—for building security tooling, guardrails, and evaluation harnesses, with sound practices in version control, testing, and CI/CD. Working knowledge of GenAI/LLM and agentic design patterns and protocols (agents, MCP) and prompt-security concepts (prompt injection, jailbreaks, tool misuse). Working knowledge of Identity and Access Management and data security/privacy controls (PHI/PII, encryption). Applied understanding of the NIST AI RMF and OWASP Top 10 for LLM.
  
  Strong written and verbal communication, including explaining AI security risks to non-technical stakeholders.

• Bachelor's degree in Computer Science, Cyber Security, Software Engineering, Artificial Intelligence, or a related field (Master's a plus).
• Strong Python/scripting and general programming ability with sound software-engineering practices (version control, testing, CI/CD). Knowledge of GenAI/LLM and agentic systems, secure MCP configuration, Application/API security, and data protection. Cloud environment (Azure & AWS) security fundamentals.

• 3+ years (mid-level) in software/security engineering, AI/ML, or platform engineering in a large enterprise setting, with a strong track record building automation and tooling.
• Hands-on Python development and CI/CD integration. Identity and Access Management fundamentals. Applied use of the NIST AI RMF and OWASP Top 10 for LLM.

Work primarily in a climate-controlled environment with minimal safety/health hazard potential.

Work primarily in a climate-controlled environment with minimal safety/health hazard potential.

This position is classified as "in-office." As an in-office role, it requires employees to work from a designated Carnival office in South Florida Monday through Thursday each week. Employees may work from their homes on Fridays.…