Responsibilities
- AI Security Architecture & Assessment: Advise on and assess the security posture of AI/ML systems, including LLMs, GenAI pipelines, and model serving infrastructure — identifying vulnerabilities, attack surfaces, and gaps against industry frameworks (e.g., OWASP LLM Top 10, MITRE ATLAS).
- Threat Modeling for AI Systems: Lead threat modeling exercises specific to AI workloads, covering adversarial inputs, prompt injection, model inversion, data poisoning, and supply chain risks across SaaS, self-hosted, and local AI deployments.
- Secure AI Integration Guidance: Advise internal teams on securely integrating SaaS AI services and APIs (e.g., OpenAI, Azure OpenAI, Bedrock) into enterprise applications, including safe handling of credentials, outputs, and user data.
- Data Security & Privacy Controls: Evaluate and recommend controls for data ingestion pipelines, RAG architectures, and vector databases to prevent unauthorized data exposure, leakage through model outputs, or non-compliant data processing.
- Collaboration & Stakeholder Engagement: Serve as a trusted security advisor bridging business stakeholders, AI/ML engineers, IT operations, and information security teams on all matters related to AI risk and security.
- Stay Current on AI Threat Landscape: Continuously track emerging AI security research, adversarial techniques, regulatory developments, and vendor security advisories to keep client guidance relevant and proactive.
- Security Documentation & Standards: Produce and maintain security architecture documentation, risk assessments, control frameworks, and guidelines tailored to the organization's AI environment.
- Strategic AI Security Roadmap: Contribute to the development of a long-term AI security strategy, including prioritized remediation roadmaps, capability maturity assessments, and investment recommendations.
- Security Awareness & Education: Develop and deliver training and awareness content for technical and non-technical stakeholders on AI-specific risks, responsible AI usage, and secure development practices for AI-powered applications.
Requirements
- 5+ years of experience in security engineering with a significant focus on cloud security and/or AppSec
- Hands-on experience implementing, managing, securing, and supporting Agentic AI solutions within an enterprise context
- Familiarity with major cloud service provider AI-focused services such as AWS Bedrock, AWS SageMaker, Azure AI Foundry, or Google Vertex
- Proficiency in at least one relevant programming language, preferably Python
- Solid understanding of generative AI concepts, Large Language Models (LLMs), context engineering, agentic tool usage, and foundational AI/ML principles
- Deep knowledge and real operational experience in the usage of Agentic Coding assistants like Claude Code, Open Code, Cursor, or Codex
- Strong written and oral communication and interpersonal skills, with the ability to explain complex technical concepts to both technical and non-technical audiences
- Demonstrated experience applying security principles to AI implementations, including data protection, access controls, and threat modeling for AI systems
- Understanding of AI-specific security challenges including prompt injection, data poisoning, supply chain security, and model extraction attacks
- Embraces emerging technologies, including AI tools, to work smarter, solve problems, and drive better business outcomes
Nice to Have
- Certifications such as: AWS Certified AI Practitioner/AWS Certified Machine Learning Engineer
- Azure AI Engineer Associate
- Claude Certified Architect
- Understanding or experience with model fine-tuning techniques
- Familiarity with red teaming of agentic systems
- Experience with policy as code languages like Cedar or Rego and Infrastructure as Code (IaC) tools like AWS CloudFormation, Terraform, OpenTofu, or equivalent technologies
- Experience designing and implementing agentic AI architectures that balance security and autonomy
- Familiarity with MCP client/server architecture vs. agentic skills and the associated security risks of each
