Forward Deployed AI Security Engineer

The Forward Deployed AI Security Engineer is a customer-facing technical operator who embeds directly in client environments to assess and secure generative and agentic AI systems at runtime. This role involves designing, deploying, tuning, and operationalizing AI security controls that ensure resilience under production loads.

You will install, configure, integrate, and defend tools in live estates, including Palo Alto Prisma AIRS, Koi Agentic Endpoint Security, Onyx, Zenity, Zscaler AI Security, and CrowdStrike Falcon AIDR. Your responsibilities include writing assessment reports, briefing executives, training operators, and ensuring the control plane functions effectively—not just until a workshop ends.

Exceptional written and spoken English is required, as you will produce precise, defensible, and unambiguous documentation, including assessment reports, architecture decisions, runbooks, and executive briefings.

• Perform AI runtime assessments. Inventory AI applications, models, agents, copilots, MCP servers, plugins, and data flows. Evaluate prompt/response paths, tool-calling behavior, identity chains, and attack surfaces. Produce scored findings covering prompt injection, jailbreaks, data leakage, model misuse, shadow AI, and unsafe autonomous actions.
• Deploy and operationalize AI security tools. Stand up, integrate, and harden Palo Alto Prisma AIRS, Koi Agentic Endpoint Security, Onyx Secure AI Control Plane, Zenity, Zscaler AI Security, and CrowdStrike Falcon AIDR.
• Design the control plane. Map traffic through AI gateways, inline intercepts, endpoint agents, SaaS connectors, and identity providers. Define allow/block/human-in-the-loop policies for prompts, tool use, MCP calls, package installs, and agent actions.
• Integrate with the existing stack. Wire AI security telemetry into SIEM/SOAR, EDR/XDR, SSE/Zero Trust, IAM, CSPM, and ticketing systems. Correlate AI events with endpoint, identity, and network signals for actionable detections.
• Red team and validate. Conduct adversarial testing against LLMs, agents, and copilots. Convert findings into enforceable guardrails and re-test to prove residual risk is managed.
• Own customer outcomes. Embed with the account to drive architecture workshops, implementation sprints, policy workshops, and knowledge transfer. Leave behind runbooks, detection content, and an operating model the customer can sustain independently.
• Communicate with precision. Write assessment reports, risk memos, architecture decision records, and board-ready summaries. Brief CISOs, AI platform owners, legal/privacy teams, and engineers in clear, concise English.
• Feed product and practice. Return field patterns, integration gaps, and detection ideas to product, threat research, and delivery teams. Help standardize playbooks for repeatable AI security deployments.

• Exceptional written and spoken English. Ability to produce 20-page technical assessments, one-page CISO briefs, and customer emails with equal clarity. Grammar, tone, structure, and technical precision are mandatory.
• Hands-on deployment experience with Palo Alto Prisma AIRS, Koi Agentic Endpoint Security, Onyx Security, Zenity, Zscaler AI Security, and CrowdStrike in production or customer environments.
• Demonstrated ability to perform AI runtime assessments, including discovering AI assets, inspecting prompt/tool-call traffic, and identifying prompt injection, jailbreaks, data exfiltration, and agent-abuse paths. Translate findings into actionable controls.
• 5+ years in cybersecurity engineering, security architecture, or technical consulting, with at least 2 years focused on cloud, application, or AI/LLM security.
• Working fluency with modern AI architecture: LLMs, SLMs, RAG, agents, MCP servers, copilots (e.g., Microsoft 365 Copilot, GitHub Copilot), model gateways, and cloud AI platforms (AWS Bedrock, Azure AI Foundry, Google Vertex AI).
• Strong integration skills across identity (Okta, Entra ID), endpoints, network/SSE, APIs, Kubernetes, and logging pipelines.
• Ability to operate independently on-site, debugging connectors, certificates, proxies, agents, and policy conflicts without back-office support.
• Willingness to travel extensively and embed with customers for multi-day or multi-week deployments.
• Bachelor’s degree in Computer Science, Cybersecurity, Engineering, or equivalent practical experience.

• Prior forward-deployed, professional-services, or resident-engineer experience (e.g., Palantir-style FDE, vendor PS, or Big Four cyber delivery).
• Vendor certifications or deep product credentials in Palo Alto Networks, Zscaler, CrowdStrike, or adjacent SSE/XDR platforms.
• Experience with AI red teaming methodologies, OWASP Top 10 for LLM/Agentic Applications, MITRE ATLAS, and NIST AI RMF.
• Scripting and automation in Python and one of Bash, PowerShell, or Go. Comfort with application and infrastructure-as-code.
• Familiarity with DLP, DSPM, secrets detection, model scanning, and software supply-chain controls for AI artifacts.
• Experience in regulated industries (financial services, healthcare, government) where AI use must be auditable.
• Public speaking, workshop facilitation, or published technical writing on AI security.

• Palo Alto Prisma AIRS: AI runtime security (network/API intercept), model security, agent security, AI red teaming, posture management, and AI gateway controls.
• Koi Agentic Endpoint Security: Discovery and risk scoring of endpoint-resident AI software, packages, extensions, local models, agents, and MCP servers; install governance and session-aware AIDR.
• Onyx: Secure AI control plane — continuous AI asset discovery, session capture, identity-aware attribution, and Guardian Agent enforcement (allow, redirect, require approval, or block).
• Zenity: AI agent observability, AI security posture management (AISPM), exposure management, runtime boundaries, and coverage across SaaS copilots, custom agents, and coding agents.
• Zscaler: Zero Trust access to GenAI, AI asset management, AI Guard runtime inspection of prompts and responses, and correlation of AI-event telemetry with the Zero Trust Exchange.
• CrowdStrike: Falcon AIDR for runtime detection and response across endpoint, SaaS, and cloud AI use; integration with Falcon XDR / Next-Gen SIEM for investigation and automated response.