Agentic AI Security Specialist

1. Agentic Threat Modeling & Secure Engineering

• Perform specialized Threat Modeling on multi-agent orchestrations, focusing on tool-use vulnerabilities, autonomous execution manipulation, goal-jacking, and cascading agent failures.
• Architect secure execution environments (Security Engineering) and sandboxes to isolate autonomous agents, ensuring strict enforcement of the principle of least privilege.
• Implement real-time proxy guardrails and Advanced Threat Protection boundaries to intercept malicious prompts, indirect prompt injections, and data exfiltration vectors.

2. Offensive Simulation & Detection Engineering

• Execute Offensive Security Testing specifically mimicking adversarial machine learning tactics against agent reasoning loops, vector databases, and memory layers.
• Design custom Detection Engineering logic to identify anomalous agent behaviors, API misuses, and malicious intent vectors before downstream execution occurs.
• Develop automated orchestration pathways (Response Engineering) to instantly isolate, throttle, or terminate rogue agent threads without disrupting core systems.

3. Core Language & Automation Engineering

• Write production-grade security proxies, wrapper interfaces, and low-latency validation controls using Python and the Go Programming Language.

Secondary Roles & Responsibilities

• Conduct proactive Cyber Threat Hunting exercises to uncover hidden signs of systemic poisoning, persistent agent manipulation, or silent model exploitation.
• Perform advanced Digital Forensics on complex agent memory logs, tracking step-by-step reasoning patterns and API execution loops following anomalous events.

2. Governance, Delivery, & Enablement

• Embed security guardrails smoothly into production deployment loops using DevOps pipelines (CI/CD) and automated testing frameworks.
• Operate within high-velocity teams using Agile Methodology & Tools (e.g., Jira, Confluence) to cleanly deliver tasks inside sprint timelines.
• Maintain a mindset of Customer Centricity, ensuring that heavy security layers and model guardrails do not negatively disrupt the user experience.
• Commitment to Continuous Learning, tracking the fast-evolving landscape of agentic frameworks (e.g., LangChain, AutoGen, CrewAI) and modern security benchmarks (MITRE ATLAS, OWASP Top 10 for LLMs).

#J-18808-Ljbffr