Researcher, Safety Training

Skills

About the Role

You will research safety post-training methods like RLHF, RLAIF, DPO and PPO, implement training interventions, build evaluations and measurement infrastructure, analyze model failures, and partner with domain experts and cross-functional teams to ship safety improvements in OpenAI products.

Requirements

• Hands-on experience training, fine-tuning or post-training language models, multimodal models or agentic systems
• Experience with RL, RLHF, RLAIF, DPO, PPO, reward modeling or preference optimization
• Strong research judgment and engineering ability
• PhD in machine learning, computer science or related field, or equivalent practical experience

Responsibilities

• Conduct research on safety post-training methods
• Develop and implement model training interventions
• Build evaluations and measurement infrastructure for safety harms
• Translate ambiguous safety problems into concrete research plans, datasets, training runs and evals
• Analyze model failures such as hallucination, prompt injection and sycophancy
• Partner with domain experts on safety training approaches
• Collaborate with research and deployment teams to incorporate safety improvements into training and launches

Benefits

• Equity
• Performance-related bonus
• Medical, dental, and vision insurance with HSA employer contributions
• Health FSA, Dependent Care FSA, and commuter pre-tax accounts
• 401(k) with employer match
• Paid parental leave, medical and caregiver leave
• Flexible PTO
• 13+ paid company holidays
• Mental health and wellness support
• Employer-paid life and disability coverage
• Annual learning and development stipend
• Daily meals in offices and meal delivery credits
• Relocation support
• Charitable donation matching and wellness stipends