Research Engineer, Safety & Alignment

About the role

Cognition is an applied AI lab building end-to-end software agents, behind Devin, the first AI software engineer. As a founding member of the safety team reporting to the Head of Safety, you will build the evaluations and red-teaming that gate every release, develop alignment methods that shape how models are trained, and work directly with the agent team on how Devin plans, acts and asks for help. This is a hands-on research engineering role.

Key responsibilities

• Design and run evaluations for failure modes such as unsafe actions, prompt injection, data exfiltration, sandbox escape, reward hacking and misuse, fast enough for every model and product release.
• Red-team Devin and its harness, find failures early and turn them into fixes and regression tests.
• Develop alignment methods such as reward modeling, preference data and constitutional approaches.
• Partner with the agent team on permissions, oversight and escalation, and on how the agent explains itself.
• Publish and share safety work through papers, evaluations and open methods.

Required profile

• Safety or alignment research experience, such as evaluations, red-teaming, alignment or interpretability at a frontier AI lab or in published research.
• Understanding of how agents fail in practice, with concrete ideas about how to fix them.
• Strong engineering fundamentals and empirical rigor in designing clean experiments and reporting results honestly.
• Comfort with adversarial thinking and enjoyment of breaking systems.
• Relevant industry experience at a frontier AI lab, applied AI company or developer tools company.
• Advanced degree (PhD in CS, ML, Physics, Mathematics or related) or equivalent industry research experience.

Required skills

• Proficiency in Python and PyTorch (or JAX)
• Ability to build eval infrastructure, run experiments at scale and read training and inference code