Applied AI Safety Engineer

Sourced from diversityjobs.com

• Develop comprehensive threat models and harm taxonomies for AI systems.
• Design and execute adversarial evaluations including red teaming and automated attacks.
• Construct reusable Python evaluation pipelines and LLM-as-a-judge workflows.
• Validate evaluators against human labels and quantify safety-utility trade-offs.
• Translate findings into actionable mitigations such as policy adjustments and prompt engineering.

• Experience delivering safety evaluations or implementing mitigations for real-world AI/ML products.
• Strong coding proficiency in Python and SQL for data analysis and code assessment.
• Proven experience designing adversarial tests and developing evaluation datasets.
• Ability to work autonomously on ambiguous safety problems.