Key Responsibilities
AI Security Leadership
- Lead the AI Security function, including development and delivery of AI Security Service Offerings, reference architectures, operational runbooks, and a roadmap for continual improvement.
- Partner with Cybersecurity Architects and Engineers, IT Infrastructure Engineers, and Application and AI Agent Developers to embed security into AI solutions from design through production.
- Partner with Project and Program Management to ensure that we’re executing the right work with the right priorities.
- Partner with GRC to keep AI use compliant with CRC policy and regulatory obligations (including NYDFS, among others), and to remediate any findings or observations.
- Maintain the ability to quickly report on current state, risk posture, and progress maturing this space.
Agent Harness Security Patterns
- Develop, publish, and maintain secure design patterns for AI agent harnesses, covering tool permissioning, sandboxing, human-in-the-loop approvals, memory and context handling, and safe failure modes.
- Define guardrails against prompt injection (direct and indirect), tool misuse, excessive agency, data leakage, and model or supply-chain compromise, aligned to OWASP Top 10 for LLM Applications, OWASP Agentic AI guidance, and MITRE ATLAS.
- Secure agents built on Microsoft Foundry (including Foundry Agent Service), and establish equivalent patterns for third-party harnesses such as Anthropic’s Claude Agent SDK.
AI Gateway & API Security
- Engineer and operate Azure APIM and AI Gateway capabilities as the policy enforcement point for model, agent, MCP, and API traffic, including authentication, authorization, rate and token limits, content safety, logging, and routing.
- Protect APIs consumed and exposed by agents with strong identity, schema validation, threat protection, and least-privilege access.
- Integrate gateway telemetry with Microsoft Sentinel and Defender for Cloud for monitoring, detection, and response.
MCP & Agent-to-Agent (A2A) Security
- Define and enforce standards for building, registering, and consuming MCP servers, including OAuth-based authorization, server allow-listing, tool and scope minimization, and supply-chain vetting.
- Establish trust, authentication, and authorization models for A2A interactions, including agent identity, delegation boundaries, and auditability of multi-agent workflows.
- Maintain an inventory of agents, MCP servers, and their connections so that exposure and blast radius are always known.
Identity & Secure Access
- Design secure access to agents, MCP servers, APIs, and data using Microsoft Entra ID, managed identities, workload and agent identities, Conditional Access, and on-behalf-of flows.
- Partner with the Data Security team to ensure agents only reach data they are entitled to, leveraging Microsoft Purview classification, DLP, and DSPM for AI.
Control Effectiveness & Assurance
- Define measurable control objectives and KPIs, and continuously validate that AI security controls are working as intended.
- Lead AI red teaming, adversarial testing, and threat modeling of agentic systems before and after production release.
- Build detections and automated response for AI-specific threats, and continuously improve controls based on findings, incidents, and emerging threats.
Qualifications
Required:
- 5+ years in Cybersecurity, with hands-on keyboard engineering and operational responsibilities.
- Demonstrated experience developing security patterns for AI agent harnesses and their supporting components, including API and AI gateways.
- Hands-on experience securing AI workloads in Microsoft Azure, including Microsoft Foundry.
- Hands-on experience engineering and operating Azure API Management (APIM), including policy authoring.
- Practical experience designing and validating security controls for AI agents, MCP servers, and A2A interactions.
- Strong understanding of identity and access patterns for workloads and agents (OAuth 2.x, OIDC, Microsoft Entra ID, managed identities).
- Working knowledge of AI-specific threats and frameworks such as OWASP Top 10 for LLM Applications, MITRE ATLAS, and NIST AI RMF.
- Ability to communicate clearly with technical and executive stakeholders, adapting messaging to the audience.
Preferred:
- Experience with Anthropic’s agent harness and related patterns (e.g., Claude Agent SDK, Claude Code, hooks, permissioning, and sub-agents).
- Experience with Azure AI Content Safety / Prompt Shields, Defender for Cloud AI threat protection, and Microsoft Purview DSPM for AI.
- Scripting or development proficiency in Powershell, Python, or other scripting, plus infrastructure-as-code (Terraform or Bicep) and KQL.
- Familiarity or experience with AI red teaming tools such as PyRIT, Garak, or similar.
- Microsoft certifications such as AZ-500, SC-100, or AI-102, or equivalent; AI security credentials (e.g., GIAC GOAA) a plus.
- Familiarity with regulatory and governance frameworks such as NYDFS 23 NYCRR 500, ISO/IEC 42001, and NIST 800-53.
- Experience in insurance or financial services.
Work Environment
- Hybrid or on-site depending on organizational needs.
- On-call rotation may be required for critical incident response.
- Required flexibility to work nights, weekends and/or holiday shifts in the event of an incident response emergency.
Location: The preferred location is Charlotte, NC. However, we will consider candidates based in other locations within the Eastern time zone.
CRC Group supports a diverse workforce and is an Equal Opportunity Employer that does not discriminate against individuals on the basis of race, gender, color, religion, citizenship or national origin, age, sexual orientation, gender identity, disability, veteran status or other classification protected by law. CRC Group is a Drug Free Workplace.