remote
AI Red Teamer
AI Red Teamer
As an AI Red Teamer at HiddenLayer, you will evaluate the security of AI systems, focusing on both predictive and generative AI models. You will identify vulnerabilities, simulate adversarial attacks, and provide actionable recommendations to improve AI system security. This role involves conducting penetration testing, developing and executing adversarial attacks, and collaborating with various teams to implement novel attacks and report on findings.
About the role
ABOUT THE ROLE:
As an AI Red Teamer at HiddenLayer, you will play a pivotal role in the ML Threat Operations group. In this role, you will evaluate the security of AI systems, focusing on both predictive and generative AI models. You will identify vulnerabilities, simulate adversarial attacks, and provide actionable recommendations to improve the security of AI systems. The ideal candidate is a proactive problem solver with hands-on experience in AI security testing and a deep understanding of machine learning models and adversarial techniques.
WHAT YOU’LL DO:
- Conduct end to end penetration testing on AI systems, with a focus on predictive and generative AI models.
- Develop and execute adversarial attacks (e.g., evasion, poisoning, and inference attacks) to identify weaknesses in predictive models.
- Develop and execute adversarial attacks (e.g., jailbreak, hallucination, context leakage, etc.) to identify weaknesses in generative AI models and applications built on top of them.
- Collaborate with data scientists, engineering, and research teams to design and implement novel attacks and relate them back to actionable recommendations.
- Stay current with the latest AI security research, trends, and adversarial tactics.
- Produce detailed reports outlining vulnerabilities, risks, and actionable recommendations.
- Contribute to the development of internal tools and frameworks for AI red teaming.
WHO YOU ARE:
- 3+ years of experience in penetration testing, with at least 1 year focused on AI systems.
- Deep understanding of attack techniques specific to machine learning and artificial intelligence systems (data poisoning, inference attacks, model injection, prompt injection, jailbreaking, etc.).
- Hands-on experience with adversarial machine learning techniques and tools (e.g., Foolbox, CleverHans, ART, Purple Llama, Garak, or proprietary solutions).
- Excellent communication skills with the ability to articulate complex concepts clearly to both technical and non-technical audiences.
- Understanding of machine learning concepts and algorithms.
- Strong problem-solving skills and the ability to think creatively to anticipate potential attack vectors.
- Proficiency in programming languages such as Python, and experience with AI frameworks like TensorFlow, PyTorch, or Keras.