AI Offensive Security Engineer

eBay•Austin, TX

3d•$136,000 - $228,600•Remote

About The Position

eBay seeks an AI Offensive Security Engineer to proactively identify and exploit vulnerabilities in AI/ML systems before adversaries do. This is a hands-on technical role focused on offensive research, adversarial attack simulation, and building the tools and automation that scale our AI security testing program. If you're an experienced offensive security engineer ready to apply your skills to one of the fastest-growing attack surfaces in the industry, this role is for you.

Requirements

5+ years in offensive security, penetration testing, or red teaming
Strong software engineering skills — you write production-quality tools, not just scripts (Python required; Java, or JavaScript a plus)
Background in traditional appsec/web pentesting (Burp Suite, fuzzing, auth bypass) — many AI vulnerabilities are classic vulns in new wrappers
Cloud security experience across AWS, GCP, or Azure — including container/Kubernetes environments and cloud-hosted AI services
Ability to independently scope, execute, and report on offensive engagements
Strong written communication — clear, technically precise findings reports

Nice To Haves

Hands-on experience attacking LLMs/GenAI systems, ML pipelines, APIs, or cloud-hosted AI services
Knowledge of common failure modes of AI agents and LLMs, risks and exploitation techniques
Experience with agentic AI frameworks, technologies and their attack surfaces
Familiarity with MLOps tooling, model serving infrastructure, and vector databases

Responsibilities

Design and execute adversarial attacks against production AI/ML systems — including prompt injection and traditional vulnerabilities
Lead AI-focused red team engagements: threat model targets, develop attack chains, execute, and deliver actionable findings
Build, maintain, and document offensive security tooling for AI system testing — attack harnesses, fuzzing frameworks, automated vulnerability scanners, and agentic attack simulations
Discover novel attack techniques against LLMs, multimodal models, agentic pipelines, and AI-integrated applications; contribute findings to the broader security community
Develop proof-of-concept exploits for discovered vulnerabilities and work with engineering teams on remediation validation
Integrate offensive testing into pipelines and security review workflows so AI red teaming is continuous, not point-in-time
Document TTPs, author internal playbooks, and mentor other team members on AI attack techniques