🛠 Tech Stack

💼 About This Role

You'll develop RLHF and post-training methods for multimodal AI systems at OpenAI. Your work will directly shape how models learn from feedback over time, enabling personalized and context-aware behavior. This is a product-grounded research role with access to frontier infrastructure and safety teams.

🎯 What You'll Do

• Develop RLHF and post-training methods for multimodal models.
• Build reward models and preference-learning pipelines for adaptive behavior.
• Design datasets, rubrics, and evaluation frameworks for user preferences.
• Collaborate with safety researchers on aligned personalization.

📋 Requirements

• 3+ years experience in machine learning research with RLHF or post-training.
• Experience with reward modeling, preference optimization, or reinforcement learning.
• Proven track record of designing experiments and reliable evaluations.
• Experience building datasets or eval pipelines grounded in human preferences.

✨ Nice to Have

• Experience with multimodal AI systems.
• Background in personalization, memory, or recommender systems.
• Familiarity with long-horizon evaluation or user modeling.

🎁 Benefits & Perks

• 💰 Competitive compensation ($380K – $445K + equity)
• 🏖️ Flexible PTO and hybrid work model (4 days in office)
• ✈️ Relocation assistance available
• 🧠 Access to frontier AI research and compute resources
• 🏥 Comprehensive health benefits