Principal AI Security Expert

Uney GmbH seeks a Principal AI Security Expert to design secure AI architectures, lead red teaming efforts, and mentor teams in Zurich. This on-site role requires deep expertise in adversarial machine learning and cybersecurity.

The company is looking for a senior specialist to define security strategies for artificial intelligence systems. The position focuses on guiding technical directions, conducting thorough security reviews during development, and mentoring other team members to ensure robust AI safety.

Responsibilities

• Create security frameworks for AI systems, including threat modeling and risk mitigation strategies.
• Oversee security audits and reviews of AI models throughout their development lifecycle.
• Work with engineering, product, and data science teams to embed security into development workflows.
• Build and maintain tools for adversarial testing to proactively identify system weaknesses.
• Direct red teaming exercises to find and exploit vulnerabilities in AI models.
• Research new attack vectors targeting classical machine learning and Transformer architectures.
• Assess AI system vulnerabilities and propose solutions to prevent exploitation.
• Study and apply methods to improve model resilience against adversarial inputs.
• Implement defenses against evasion, poisoning, extraction, and membership inference attacks.
• Establish guardrails to stop harmful outputs and unsafe AI behaviors.
• Set up input validation and monitoring systems to detect anomalous model activity.
• Develop protections against prompt injection, context confusion, and semantic attacks.

Requirements

• PhD in Computer Science, Cybersecurity, AI, or a related field with research in AI security, or a Master’s degree plus 10+ years of professional experience.
• At least 10 years of professional experience in cybersecurity, machine learning, or AI security.
• Minimum 8 years in a senior or lead role with proven technical leadership.
• Demonstrated history of finding and fixing critical security issues in AI systems.
• Published research or significant contributions to the field of AI security.
• Deep knowledge of adversarial machine learning and attack methods for ML and Transformers.
• Practical experience with adversarial attack frameworks such as FGSM, PGD, and C&W.
• Proficiency in Python and deep learning libraries like PyTorch and TensorFlow.
• Strong grasp of cryptography, secure computation, and privacy-preserving techniques.
• Understanding of prompt injection, jailbreak, and context confusion attack vectors.
• Experience with security testing tools, vulnerability assessment, and risk mitigation.