As Principal Architect for AI Security, you will define and lead the security architecture strategy for our AI platform, products, and infrastructure. This is a rare opportunity to shape how a fast-moving organization thinks about security at the intersection of machine learning, cloud infrastructure, and product delivery. You will operate as a technical authority — influencing roadmap, mentoring senior engineers, and partnering with leadership to ensure our AI systems are trustworthy, resilient, and defensible.
Compensation
This position pays between $148,800 - 223,200 based on experience and geographical location.
Responsibilities
Define the end-to-end security architecture for AI/ML systems, including training pipelines, inference infrastructure, model storage, and APIs.
Develop and maintain a threat model for AI-specific attack surfaces: model poisoning, adversarial inputs, prompt injection, model inversion, and supply chain risks.
Lead design reviews for high-impact systems, setting standards for secure-by-default AI development.
Own the AI security roadmap, aligning it with product goals, compliance requirements, and industry best practices.
Serve as an authority on AI security decisions across engineering, product, and research teams.
Requirements
Nice to have: NIST AI RMF
Nice to have: CISSP
Nice to have: CCSP
Nice to have: Cloud security specializations
Nice to have: Published research in AI/ML security
Lead deployment, administration, and strategic use of the HiddenLayer platform to protect AI/ML systems from adversarial attacks and other AI-specific threats.
Responsibilities
Deploy, configure, and administer the HiddenLayer platform
Integrate HiddenLayer with enterprise SIEM, SOAR, EDR, vulnerability management, and cloud security platforms
Create detection rules, dashboards, and executive reporting on AI security posture
Protect AI models against adversarial attacks, model theft, prompt injection, model poisoning, and unauthorized inference
Develop monitoring and detection strategies for production AI workloads