Senior AI Security Engineer

Plaster Group, LLC; Seattle (WA); On-site; USD 150,000 - 190,000; Full time

Job summary

Plaster Group, LLC is seeking an experienced security engineer to safeguard AI workloads and reduce AI risk by strengthening infrastructure hygiene. You will design and operate security controls for foundation models, integrate zero-trust measures, and work with enterprise AI gateways across multiple platforms including Claude, ChatGPT, and Copilot.

The role emphasizes hands-on deployment of AI gateways, guardrails for prompt injection, data leakage, and governance across MCP servers and

Qualifications

• 7+ years in security engineering, with 2+ years securing AI, ML, or generative AI systems in production.
• Hands-on experience deploying or operating an AI gateway, LLM proxy, or equivalent centralized AI control plane
• Demonstrated understanding of LLM and agent attack surface: prompt injection, data leakage through model output, insecure tool use, supply chain risk in models and dependencies
• Practical familiarity with OWASP Top 10 for LLM Applications, MITRE ATLAS, and NIST AI RMF
• Strong cloud security background, Azure preferred, including identity, network, and workload controls
• Python proficiency and comfort working with model APIs, SDKs, and evaluation tooling
• Experience partnering with engineering and data teams as an advisor rather than a gatekeeper

Responsibilities

• Design, implement, and operate security controls for foundation model usage across Claude, ChatGPT, and Microsoft Copilot
• Support deployment and hardening of the enterprise AI gateway, including model allowlists, per-team keys, rate and budget limits, and centralized audit logging
• Engineer input and output guardrails covering prompt injection, sensitive data egress, credential and secret leakage, and unsafe output
• Integrate inline DLP and content inspection at the gateway so AI traffic is subject to the same data controls as other egress paths
• Establish security requirements for RAG architectures, including source grounding, index access control, and prevention of oversharing through model responses
• Define and enforce governance for MCP servers and agent tooling, including registry, approval, and runtime policy enforcement
• Implement identity and authorization patterns for agents and other non-human identities: OAuth-based access, token lifecycle management, scoped tool permissions, and least privilege
• Assess and mitigate agentic risk classes including excessive agency, capability chaining, tool and memory poisoning, and unsafe autonomous action
• Extend zero trust controls to AI workloads across identity, device, network, application, and data pillars
• Partner with identity engineering on conditional access, workload identity, and privileged access for AI systems and service principals
• Reduce AI risk through infrastructure hygiene: configuration baselines, egress control, secrets management, and dependency and supply chain integrity
• Build monitoring and detection content for AI systems covering prompt behavior, tool invocation patterns, anomalous output, and data movement
• Conduct AI red team and adversarial testing, including direct and indirect prompt injection, jailbreak, and data extraction scenarios
• Conduct security reviews of AI applications, third-party AI vendors, and AI features in existing SaaS
• Map controls to recognized frameworks including the OWASP Top 10 for LLM Applications, MITRE ATLAS, and the NIST AI Risk Management Framework
• Contribute to AI security standards, acceptable use guidance, architecture review criteria, and enablement material for engineering teams

Skills

Security engineering

AI security

Python

Cloud security

Identity & access management

Tools

AI gateway

LLM proxy

Model APIs

SDKs

Job description

POSITION SUMMARY

Safeguard AI workloads across the foundation and reduce AI risk through stronger infrastructure hygiene, in partnership with zero trust engineering.

RESPONSIBILITIES

• Design, implement, and operate security controls for foundation model usage across Claude, ChatGPT, and Microsoft Copilot
• Support deployment and hardening of the enterprise AI gateway, including model allowlists, per-team keys, rate and budget limits, and centralized audit logging
• Engineer input and output guardrails covering prompt injection, sensitive data egress, credential and secret leakage, and unsafe output
• Integrate inline DLP and content inspection at the gateway so AI traffic is subject to the same data controls as other egress paths
• Establish security requirements for RAG architectures, including source grounding, index access control, and prevention of oversharing through model responses

Agentic and MCP Security

• Define and enforce governance for MCP servers and agent tooling, including registry, approval, and runtime policy enforcement
• Implement identity and authorization patterns for agents and other non-human identities: OAuth-based access, token lifecycle management, scoped tool permissions, and least privilege
• Assess and mitigate agentic risk classes including excessive agency, capability chaining, tool and memory poisoning, and unsafe autonomous action

Zero Trust Integration

• Extend zero trust controls to AI workloads across identity, device, network, application, and data pillars
• Partner with identity engineering on conditional access, workload identity, and privileged access for AI systems and service principals
• Reduce AI risk through infrastructure hygiene: configuration baselines, egress control, secrets management, and dependency and supply chain integrity

Detection, Testing and Assurance

• Build monitoring and detection content for AI systems covering prompt behavior, tool invocation patterns, anomalous output, and data movement
• Conduct AI red team and adversarial testing, including direct and indirect prompt injection, jailbreak, and data extraction scenarios
• Conduct security reviews of AI applications, third-party AI vendors, and AI features in existing SaaS

Governance

• Map controls to recognized frameworks including the OWASP Top 10 for LLM Applications, MITRE ATLAS, and the NIST AI Risk Management Framework
• Contribute to AI security standards, acceptable use guidance, architecture review criteria, and enablement material for engineering teams

REQUIRED QUALIFICATIONS

• 7+ years in security engineering, with 2+ years securing AI, ML, or generative AI systems in production
• Hands-on experience deploying or operating an AI gateway, LLM proxy, or equivalent centralized AI control plane
• Demonstrated understanding of LLM and agent attack surface: prompt injection, data leakage through model output, insecure tool use, supply chain risk in models and dependencies
• Practical familiarity with OWASP Top 10 for LLM Applications, MITRE ATLAS, and NIST AI RMF
• Strong cloud security background, Azure preferred, including identity, network, and workload controls
• Python proficiency and comfort working with model APIs, SDKs, and evaluation tooling
• Experience partnering with engineering and data teams as an advisor rather than a gatekeeper

PREFERRED QUALIFICATIONS

• Experience with Microsoft 365 Copilot security posture, including oversharing remediation and sensitivity label enforcement
• Familiarity with AI red team tooling such as PyRIT or Garak
• Experience with MCP architecture and tool authorization patterns
• Experience with Microsoft Defender for Cloud Apps, Purview, or an equivalent DSPM platform in an AI context
• Working knowledge of ISO/IEC 42001 or the EU AI Act

Certifications

CISSP, AZ-500, SC-100, CCSP