Research Engineer, Safety & Alignment

What you'll do

• Design and run safety evaluations for autonomous-agent failure modes; build regression tests from red-teaming findings.
• Develop alignment methods with the post-training team, including reward modeling and preference data.
• Partner with the agent team on permissions, oversight, escalation, and explanations.

What they ask for

• Hands-on safety or alignment research experience, demonstrated through evaluations, red-teaming, alignment, or interpretability work.
• Understanding of agent failure modes and how to measure them.
• Proficiency in Python and PyTorch (or JAX), with ability to build evaluation infrastructure, run experiments, and read training and inference code.
• Strong empirical rigor and comfort with adversarial testing.

Nice to have

• Experience at a frontier AI lab, applied AI company, or developer tools company.
• PhD in a related discipline, or equivalent industry research experience.

Tools

Python, PyTorch, JAX

AI focus

Safety and alignment for an autonomous software agent, including evaluations, red-teaming, alignment methods, and agent oversight.