AI Security SME

Role overview

Architecture and risk

• Design and review secure architectures for ML/LLM training, fine-tuning, inference, RAG, agents, plugins, tool calling, and APIs.
• Define on-premises, hybrid, and cloud reference architectures, including Azure OpenAI, AWS Bedrock, GCP Vertex, and self-hosted models.
• Perform AI threat modeling for poisoning, model theft, prompt injection, jailbreaks, supply-chain risks, and inference attacks; use MAESTRO or similar.
• Map controls to NIST AI RMF, ISO/IEC standards, OWASP GenAI/LLM Top 10, CSA guidance, and MITRE ATLAS.

Controls and assurance

• Oversee controls for data pipelines and provenance, model and artifact integrity, access, isolation, rate limiting, abuse detection, prompt guardrails, monitoring, data classification, masking, DLP, and MCP server hardening.
• Partner with engineering and data science on robustness, observability, fallbacks, safety and quality evaluation, bias, toxicity, and hallucination monitoring.
• Contribute to SLOs/SLAs and security and reliability KPIs; embed scanning, dependency checks, policy-as-code, and red teaming into AI CI/CD.

Incident response and governance

• Develop playbooks for prompt abuse, model exfiltration, data leakage, and compromised agents; lead or support AI red/blue/purple-team exercises.
• Advise on emerging AI regulations and define policies for responsible AI, data usage, model lifecycle, and third-party AI risk.
• Run workshops, training, and awareness for engineering, security, and business teams.

Required experience and expertise

• 8–12+ years in cybersecurity, with 3–5+ years focused on AI/ML or data platforms; periods may overlap.
• Hands-on cloud platform and AI-service experience with Azure, AWS, and GCP, plus at least one agentic/GenAI stack.
• Track record securing ML pipelines, LLM/RAG systems, or agentic/automation platforms.
• Strong knowledge of cryptography, IAM, network and application security, data protection, and privacy concepts including PII, PHI, and DPIA.
• Experience mapping to standards such as NIST AI RMF, ISO/IEC 27001 and 42001, SOC 2, OWASP, MITRE ATT&CK/ATLAS, or CSA guidance.
• Excellent communication skills to explain AI risks as business and technical requirements.

Core technical skills

• AI/ML and GenAI fundamentals, including model types and training/inference flows; data pipelines, feature stores, embeddings, and vector databases.
• AI penetration testing for prompt injection, model tampering, data poisoning, output manipulation, exfiltration, shadow AI, and insecure integrations.
• Kubernetes, containers, API gateways, secrets management, zero trust, and AI-system logging, tracing, safety, quality, health, and latency monitoring.
• Strong scripting and automation for security tooling and assessments; Python is preferred.

Certifications and training

• Required: one or more core security certifications, such as CISSP, CISM, CCSP, or ISO 27001 Lead Implementer/Lead Auditor.
• Required: one or more cloud security certifications, such as CCSK, AWS Security Specialty, Azure Security Engineer, or Google Professional Cloud Security Engineer.
• Required or to be obtained within 6–12 months: AI security training/certification, including relevant NIST/ISO training, CAISP or similar, AI red-team training, or vendor AI/ML certification with security emphasis.
• Desirable: GIAC/GWAPT/GXPN/GCLD or similar; training in secure MLOps, privacy engineering, OWASP GenAI/LLM Top 10, MITRE ATLAS, or CSA MAESTRO.

About the company