Senior Applied Scientist, AI Security

Senior Applied Scientist, AI Security Department: Data Science Team: Data Scientist Location: Sunnyvale, United States Type: Full-Time **About the Role** ### **Job Description** Our mission is to make Uber the industry model for a secure and trustworthy AI ecosystem through differentiated, highly scalable, and extensible products, engineering standards, policy, and open communications. We are focusing on building robust defense mechanisms that protect the entire AI lifecycle—from internal infrastructure to external user-facing products. This includes securing the technology platform to protect model integrity, prevent data leakage, and enable safe business velocity. Lead efforts within the organization to drive the development of secure AI/ML-based solutions in support of user-facing products, internal downstream services, or infrastructure tools and platforms used across Uber. ### **About the Team** Our mission is to make Uber the industry model for AI Safety and Security through differentiated products, engineering standards, and policy. We focus on a defense-in-depth approach, building protections that span from internal developer tools to external customer agents, ensuring that our AI systems remain resilient against adversarial attacks while preserving user trust. ### **About the Role** Lead efforts to develop and evaluate security controls for large-scale machine learning models, harden retrieval-augmented generation (RAG) systems, alignment-tune large language models (LLMs), and secure agentic workflows across internal and external boundaries. This role requires a strong foundation in both traditional machine learning, advanced LLM technologies, and adversarial machine learning. **What the Candidate Will Need / Bonus Points** \-\-\-\- What the Candidate Will Do ---- 1. Develop and evaluate large-scale machine learning model systems in production with a focus on adversarial robustness, input validation, and output sanitization. 2. Propose, design, and analyze large-scale online experiments to test safety guardrails and detect ecosystem vulnerabilities. 3. Define and implement metri