Research Scientist - [Seed Model - Alignment & Preparedness]

What you'll do

• Train safer agents using RL, SFT, and preference optimization; design environments, rewards, and data to reduce unsafe behavior.
• Assess frontier risks and build safeguards, automated red teaming, monitoring, and oversight methods.
• Analyze real-world failures and collaborate across research, engineering, infrastructure, and product teams to deploy effective methods.

What they ask for

• Strong foundations in ML, deep learning, and statistics, with research or engineering experience in LLMs, RL, or related fields.
• Proficiency in Python and PyTorch (or similar), including independent algorithm implementation, experiment pipelines, and debugging.
• Hands-on experience in at least one of agent RL, model training, dataset development, agent systems, evaluation, or ML infrastructure.
• Ability to turn open-ended model-behavior questions into experiments and clearly communicate results and limitations.

Nice to have

• Experience with agent RL, SFT, RLHF/RLAIF/RLVR, distributed training, or automated red teaming.
• Research or experience in AI control, scalable oversight, CoT monitoring, mechanistic interpretability, prompt injection defense, or high-stakes risk mitigation.

Tools

Python, PyTorch, RL, SFT, RLHF, RLAIF, RLVR

AI focus

AI safety research for frontier models and agents, including agent RL, risk assessment, automated red teaming, monitoring, oversight, and deployment safeguards.