AI Red Team Specialist – Generative AI & Agentic Systems

About the role

BLEND is building a bench of offensive security specialists for an AI red‑team programme with a client. The focus is on attacking generative AI systems – models, LLM applications, and autonomous agents – and delivering actionable findings.

Key responsibilities

• Probe LLM applications and agentic systems for exploitable behaviour such as prompt injection, guardrail bypass, system‑prompt extraction, tool/function‑call abuse, data exfiltration via RAG pipelines, and privilege escalation across agent boundaries.
• Demonstrate real impact of findings, ensuring jailbreaks have measurable effect beyond curiosity.
• Document findings with reproducible steps, root‑cause analysis, severity rating, and mapping to the OWASP Top 10 for LLM Applications.
• Operate strictly within a defined scope and adhere to scope boundaries.

Required profile

• Hands‑on experience with generative AI security (professional or self‑directed) supported by verifiable evidence such as a CVE credit, bug‑bounty profile, vendor acknowledgement, published write‑up, conference talk, or certification verification.
• Fluent written English; the primary deliverable is a clear technical report.
• Willingness to complete an identity check and sign an NDA before accessing client systems.

Required skills

• Web penetration testing experience (≈2 years) or equivalent AI security work.
• Offensive security certifications (e.g., OSCP, OSWE, OSCE3, GIAC).
• Proficiency in Python or NodeJS for building custom testing harnesses.
• Active participation on platforms such as Hack The Box, TryHackMe or CTFtime.