How to apply for AI Security Research Engineer

About 0Labs

0Labs is a cutting-edge AI security company that focuses on ensuring the safe deployment of AI agents through continuous evaluation and monitoring. Their mission to bridge the gap between AI research and practical security makes them a unique and impactful place to work for those passionate about AI safety.

About the role

As an AI Security Research Engineer, you will be at the forefront of developing and stress-testing evaluation systems for AI agents, specifically focusing on purple teaming to benchmark monitoring mechanisms. Your work will directly influence how AI systems are secured against adversarial threats, making a tangible impact on the safety of AI deployments.

A typical day

A typical day might involve running adversarial evaluations on agent deployments, analyzing results to improve scoring frameworks, and collaborating with the product team to implement security fixes. You'll also spend time researching new attack surfaces and presenting findings to guide engineering decisions.

Who 0Labs is looking for

• Has hands-on experience with purple teaming or red teaming AI systems, not just traditional cybersecurity.
• Is proficient in Python and familiar with ML frameworks like PyTorch or TensorFlow, with a track record of building evaluation pipelines.
• Can demonstrate translating research findings into engineering solutions, such as published papers or implemented security tools.
• Understands agent architectures and attack surfaces specific to LLM-based agents, including prompt injection and jailbreaking.

Tips for this application

• Highlight any experience with continuous evaluation systems or benchmarks for AI safety, such as contributions to frameworks like AI Control or similar.
• Include specific examples of adversarial evaluations you've run, detailing the attack surface and the scoring framework used.
• Mention any open-source projects or tools you've developed for AI security testing.
• Tailor your resume to emphasize both research and engineering aspects, showing you can bridge the gap.
• If you have a portfolio, include a link to a GitHub repo or blog post demonstrating your AI security work.

What to cover in your cover letter

['Emphasize your passion for AI safety and your understanding of the importance of continuous evaluation.', 'Describe a specific project where you conducted adversarial testing on an AI system and how it led to improvements.', 'Explain how you approach translating research into engineering decisions, using a concrete example.', "Mention your familiarity with 0Labs' mission and how your skills align with their focus on agent security."]

Research before applying

• Read 0Labs' blog or any published research on AI control and agent evaluation.
• Familiarize yourself with the concept of 'purple teaming' in AI security and how it differs from traditional approaches.
• Look into the AI control evaluation harness and understand its current capabilities and limitations.
• Study recent adversarial attacks on LLM agents, such as prompt injection and data poisoning.

Likely interview topics

Based on the job description, expect questions about:

• How would you design a purple teaming evaluation for a multi-agent system?
• Describe a time you identified a vulnerability in an AI system and how you mitigated it.
• What metrics would you use to score the effectiveness of a monitoring mechanism for agent behavior?
• Explain the difference between red teaming and purple teaming in the context of AI security.
• How do you stay updated on the latest AI security research and apply it to product development?

Common mistakes to avoid

• Don't focus solely on traditional cybersecurity experience without linking it to AI-specific threats.
• Avoid vague statements like 'I'm passionate about AI safety' without concrete examples.
• Don't neglect the engineering aspect; this role requires building tools, not just theoretical research.