Research Engineer, Safety & Alignment

📋 Description

• Design and run evaluations for the failure modes that matter for autonomous agents: unsafe actions
• Attack the agent and its harness systematically. Find the failures before customers do and turn
• Work with post-training on reward modeling, preference data, constitutional approaches, and other
• Partner with the agent team on permissions, oversight, and escalation: when Devin should act, when
• Contribute to Cognition's external safety work through papers, evals, and open methods where it

🎯 Requirements

• Hands-on work on evaluations, red-teaming, alignment, or interpretability at a frontier AI lab or
• You know how agents fail in practice: tool misuse, specification gaming, long-horizon drift
• Proficiency in Python and PyTorch (or JAX). You can build eval infrastructure, run experiments at
• You design clean experiments, report results honestly, and know the difference between a real
• You enjoy breaking systems and are good at it.
• Prior experience at a frontier AI lab, applied AI company, or developer tools company.