Research Engineer, Safety & Alignment

Highlights

• Build agentic safety evals that gate every model release.
• Red-team Devin to find failures before customers do.
• Shape how Devin plans, acts, and asks for help.

Ideal candidate

Empirically rigorous researcher who loves breaking agentic systems and making production agents safer. Thrives hands-on at the intersection of alignment research and real-world deployment.

Requirements

• Hands-on safety or alignment research experience with impact on training or deployment
• Understanding of agentic failure modes like tool misuse, prompt injection and drift
• Proficiency in Python and PyTorch or JAX and eval infrastructure at scale
• Empirical rigor designing clean experiments and reporting results honestly
• Comfort with adversarial thinking and systematically breaking systems
• Prior experience at frontier AI lab or applied AI company
• PhD in Computer Science, Machine Learning, Physics, Math or equivalent experience