AI Security Lead

Responsibilities - AI Security Evaluations: Conduct structured adversarial testing, vulnerability analysis, and security assessments on AI models, datasets, and pipelines to identify technical weaknesses. - Design AI Security Testing Methodologies: Develop evaluation strategies, attack simulations, and test cases that examine model behavior under adversarial and stress conditions. - Analyze Model Vulnerabilities: Review model outputs, logs, and error patterns to detect indicators of poisoning, evasion, extraction attempts, or unsafe behavior requiring mitigation. - Support Development of AI Defense Mechanisms: Contribute technical insights for designing and implementing guardrail components, safety filters, validation controls, and secure model behavior constraints. - Conduct Research on Emerging AI Threats: Study new attack surfaces such as prompt injection, model inversion, indirect influence attacks, and supply chain vulnerabilities relevant to modern AI systems. - Build Security Testing Scripts and Pipelines: Create and maintain scripts, automation tasks, and evaluation pipelines that streamline security testing, red team simulations, and model behavior monitoring. - Document AI Security Findings: Prepare technical reports, vulnerability summaries, experiment results, and evidence artifacts that support security reviews and R&D decision-making. - Collaborate with Cross Functional Teams: Work with engineering, Responsible AI, cybersecurity, and data science teams to jointly investigate issues and integrate necessary security improvements. - Secure Model Deployment Activities: Assist in validating the security posture of models before deployment by verifying guardrail integration, access controls, and vulnerability remediation. - Monitor Emerging AI Security Tools and Techniques: Continuously evaluate new defensive methods, adversarial testing libraries, and AI security research to enhance testing capabilities and inform tooling upgrades. Technical and Professional Requirements: - Strong proficiency in Python with experience building security evaluation scripts and adversarial testing utilities. - Knowledge of adversarial ML techniques including evasion, poisoning, extraction, inversion, and prompt manipulation methods. - Familiarity with adversarial testing frameworks and libraries (e.g., TextAttack, IBM ART, PyRIT, MITRE ATLAS tools). - Understanding of ML and DL model behavior, architecture vulnerabilities, and LLM‑specific attack vectors. - Experience with red‑team style testing for LLMs, RAG systems, and multimodal AI systems. - Knowledge of MLOps/LLMOps workflows and integrating security tests into CI/CD evaluation pipelines. Preferred Skills Technology->AI-Responsible AI->Responsible AI->explainable ai