Researcher, Loss of Control

📋 Description

• Design mitigation components for loss of control risk.
• Integrate safeguards across product and research with teams.
• Evaluate trade-offs in coverage, robustness, latency, and utility.
• Collaborate with risk modeling, evaluations, and policy partners.
• Execute red-teaming and testing to stress-test mitigations.
• Iterate protections as model capabilities evolve.

🎯 Requirements

• AI safety passion and real-world impact.
• Experience in deep learning and transformer models.
• Proficient with PyTorch or TensorFlow.
• Strong data structures, algorithms, and software engineering.
• LLM training/fine-tuning: distillation, supervised fine-tuning, policy optimization.
• Experience designing safeguards, control mechanisms, or monitoring systems.