Staff Security Engineer - AI Security

Rivian•Atlanta, GA

22h

About The Position

As a Security Engineer at Rivian, you will spearhead the adversarial evaluation of our AI-enabled features and internal platforms. This role will operate across Offensive Security, Secure Architecture, and AI/ML. You will act as one of Cybersecurity Assurance’s primary authorities on AI-related risk. You will independently evaluate, challenge, and validate the safety and integrity of these AI initiatives. You will both break and build, applying attacker mindset & adversarial thinking to challenge the assumptions in our entire AI ecosystem, while simultaneously creating the automated tools and guardrails needed to continuously expose and mitigate critical risks.

Requirements

Adversarial Mindset: Experience working in Offensive Security, Red Teaming, Vulnerability Research, Penetration Testing or adjacent role, with a demonstrated ability to evaluate complex systems with the mindset of a determined attacker to find meaningful business impact
AI/ML Fluency: Deep understanding of generative AI systems, including RAG pipelines, large language models, vector databases, and multi-agent orchestration frameworks, and their specific security failure modes You don't just know what "Indirect Prompt Injection" is, you know how to bypass filters to achieve it.
Engineering Proficiency: Hands-on experience building, testing, or securing applications using Python/Go. You are familiar with modern ML frameworks (like PyTorch, Hugging Face, or LangChain) and bring practical experience utilizing specialized AI red teaming tools (such as Garak or PyRIT).
Courage & Clarity: You are unafraid to challenge assumptions or expose uncomfortable truths in service of system safety, and skilled at communicating high-impact risks clearly to executives and technical peers

Nice To Haves

Adversarial Depth: Bonus points if you have contributed to the security or AI community through open-source tools, research, or vulnerability disclosures that demonstrate your depth in adversarial thinking

Responsibilities

AI Red Teaming: Design and execute deep, unconstrained adversarial assessments on our applied AI systems (e.g., voice assistants, RAG services, AI agents), testing safety boundaries and guardrails through offensive security assessments
Challenge Assumptions: Push AI guardrail mechanisms beyond their intended limits. Your goal is to determine exactly when and how these controls fail, proving the difference between a system's expected behavior and its actual resilience against an attacker
Assess Agentic Workflows: Evaluate multi-agent and LLM-integrated workflows for complex risks such as privilege escalation, unsafe action chaining, direct/indirect prompt injection, and other unintended side effects.
Automate adversarial testing: Build AI-driven frameworks to scale our Red Teaming and vulnerability discovery. You will leverage LLMs, anomaly detection, and corpora of attack prompts to enhance our automated coverage.
Build Cross-Functional Tooling: Design and operate AI-driven security agents and automated frameworks that amplify the efforts of our broader Cybersecurity organization.
Co-Design Defenses: Partner closely with engineering teams to brainstorm, review & validate remediations for the gaps you find
Scale AI Security Practices: Threat model AI architectures (LLMs, vector stores, data flows) and build frameworks for continuous AI-driven adversarial testing that scale across all product teams.
Proactive Security Research: Stay ahead of the curve. Spend time researching new AI attacks and prototyping guardrails to see if they work in our environment.
Foster Security Culture: Collaborate within a cross-functional team to develop relationships, influence senior stakeholders, and drive alignment on AI risk tolerance and mitigation priorities