Research Safety

Skills

About the Role

You will investigate how models handle harmful and dual-use requests and design experiments that guide training and evaluation. Depending on project needs, you will build filtering pipelines and classifiers, apply safety-focused post-training methods, create evaluations and synthetic data, red-team models, and design mitigations.

Requirements

• Bachelor’s degree or equivalent experience in a relevant discipline
• AI safety research experience
• Experience with RLHF, RLAIF, alignment and preference modeling, deliberative alignment, safety evaluations, or red-teaming
• Proficiency in Python
• Familiarity with PyTorch, TensorFlow, or JAX
• Ability to debug distributed training and write scalable code
• Written technical communication
• Experience with long-horizon, multi-step, or agentic evaluations
• Synthetic data generation
• Red-teaming and jailbreaking
• Knowledge of scalable oversight, reward hacking, and jailbreak robustness

Responsibilities

• Build data-filtering pipelines and quality classifiers for pre-training corpora
• Apply safety-focused post-training techniques
• Design, build, and maintain safety evaluations for long-horizon and agentic tasks
• Generate and curate synthetic data for training and evaluation
• Red-team models and products to identify failure modes, jailbreaks, and risks
• Design mitigations for identified risks

Benefits

• Health, dental, and vision benefits
• Unlimited PTO
• Paid parental leave
• Relocation support as needed