SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Armadin Security is building autonomous proactive security systems led by Kevin Mandia (founder of Mandiant, acquired by Google for $5.4B). The company brings together researchers and engineers from Google, xAI, Meta, Stanford, Georgia Tech, Waterloo, Berkeley, and MIT to reinvent security for the AI era.
As a Senior Red Team Operator, you will run hands-on offensive operations against AI systems and encode your tradecraft into AI agents that operate at machine speed. Your expertise will directly train autonomous attack agents that discover, exploit, and chain vulnerabilities across external, internal, identity, cloud, and application attack surfaces. In recent engagements, the platform generated over 17 million offensive actions in three days and validated 38 attack paths.
Key responsibilities include:
- Execute full kill chains against LLMs, agents, RAG pipelines, and surrounding applications, chaining prompt injection, tool abuse, and integration flaws into novel attack paths
- Prove exploitability by turning jailbreaks and model behaviors into reproducible exploits with clear impact evidence
- Map blast radius from poisoned documents and hijacked agents through tool calls, credentials, identity, cloud, and data access
- Abuse tool and function calling, MCP servers, plugins, memory, and multi-agent handoffs to compromise agent behavior
- Attack the full AI supply chain including training data, embeddings, vector stores, model weights, and MLOps infrastructure
- Review agent frameworks and model-serving code for injection paths and insecure design using AI-assisted analysis
- Conduct black-box probing, adversarial input generation, model extraction, and membership inference attacks
- Translate attack tradecraft into techniques and decision logic that enable autonomous agents to operate at expert level and machine speed
- Pressure-test the platform's own models and agents to identify naïveté, manipulability, and errors
- Build scripts, harnesses, and tools that automate AI attacks at scale
- Deliver clear attack narratives to engineers, executives, and legal counsel
- Research emerging AI attack classes and feed findings into the platform
- Operate with startup-level autonomy, balancing concurrent operations and internal research
Requirements:
- Deep, hands-on AI offense expertise beyond OWASP Top 10 for LLM Applications, including direct/indirect prompt injection, jailbreak chains, system prompt extraction, and data exfiltration
- Depth in at least one additional AI attack surface: agentic systems (tool/function-call abuse, MCP/plugin exploitation, memory poisoning, multi-agent trust boundaries); RAG and data pipelines (vector store/embedding attacks, retrieval poisoning, cross-tenant leakage); adversarial ML (evasion, model extraction, membership inference, training/fine-tuning data poisoning); or AI supply chain/infrastructure (malicious artifacts, unsafe serialization, model registries, inference/MLOps platforms)
- Strong application security fundamentals including web and API exploitation, authentication/authorization bypass, and injection chains
- Proficiency with tools like Burp Suite, Caido, Garak, PyRIT, promptfoo, and Adversarial Robustness Toolbox
- Systems understanding of LLMs, agent frameworks, and model-serving stacks
- Proficiency across Linux, Windows, macOS, and major cloud AI platforms with scripting in Python, Bash, or similar
- AI-forward working style with existing use of AI tools for recon, analysis, and tooling
- Sharp human-in-the-loop judgment, ability to absorb information quickly and make time-sensitive decisions
- Builder's mindset with startup-level scrappiness and self-direction
- Clear communication skills for translating complex AI attacks to engineers, executives, and legal counsel
- Must be eligible to work in the United States without sponsorship
- Ability to travel up to 20% for on-site engagements or client meetings
Preferred qualifications:
- Experience building AI systems (LLM fine-tuning, agentic frameworks like LangChain/AutoGen, ML pipelines)
- Shipped offensive AI tooling (prompt injection harnesses, jailbreak fuzzers, Burp/Caido extensions, open-source projects)
- Public AI security research (bug bounties, CVEs, AI CTF wins, conference talks at DEF CON AI Village)
- Active participation in AI security challenges (Gandalf, Crucible, HackAPrompt, PortSwigger labs)
- Familiarity with MITRE ATLAS, OWASP Top 10 for LLM Applications, NIST AI Risk Management Framework
- Relevant certifications (OSWE, OSAI, Burp Suite Certified Practitioner, GWAPT, emerging AI security credentials)
