Research Engineer / Research Scientist, Human Flourishing

OpenAI • San Francisco, CA

10d • $380,000 - $500,000

About The Position

The Human Flourishing team, part of OpenAI’s Personal AGI organization, works to ensure AI systems understand and advance human flourishing. The team measures and improves models’ understanding of people’s values and long-term goals, and develops capabilities in human domains such as education and mental health to increase the benefits people receive from ChatGPT.

The role is for a hands-on researcher who will turn difficult-to-define signals of long-term flourishing into measurable improvements in model behavior. The researcher will help set research direction, build tools and experiments to test ideas, and bring successful work into deployed models and products. The work spans research, model training, evaluation, and collaboration with product teams, engineers, partnerships teams, and domain experts. Relevant experience may include improving frontier models through pretraining or reinforcement learning/post-training, particularly on less-verifiable problems, or combining strong LLM research skills with expertise in a field such as education or mental health. Prior experience in those domains is not required.

Requirements

• Unusually strong research record in pretraining or RL/post-training research, especially in a hard-to-verify area; alternatively, strong LLM research skills combined with deep expertise in education, mental health, or a related human-benefit field.
• Ability to turn ideas into rigorous experiments, write and debug code, and work directly with data and training infrastructure.
• Ability to identify consequential problems and actionable steps toward solving them.
• Ability to take ownership of ambiguous problems and choose methods based on their potential impact.
• Focus on whether capabilities can scale, generalize, and reach real-world use, not only on demonstrating them in a paper or prototype.
• Ability to work across disciplines and identify important bottlenecks to scaling impact in multidisciplinary problems.
• Care about helping AI improve people’s lives and regard this as a responsibility associated with increasingly capable models.

Nice To Haves

• Deep experience improving frontier models through pretraining or reinforcement learning/post-training, especially on less-verifiable problems.
• Strong LLM research skills combined with deep expertise in education, mental health, or a related field of human benefit.
• Prior experience in education or mental health is not required.

Responsibilities

• Develop scalable training methods across the training stack to improve how models understand people’s values, support their agency, and help them achieve long-term goals.
• Lead high-leverage research from an open question through defining success, running experiments, and integrating improvements into frontier models.
• Build RL environments for difficult-to-verify outcomes and study whether beneficial behavior generalizes across settings.
• Create evaluations grounded in human expertise and feedback to identify capability gaps, unintended incentives in RL, and failures that aggregate scores may miss.
• Test whether measures of model behavior predict better learning, wellbeing, and other long-term outcomes, and use findings to guide training.
• Build new 0→1 capabilities in human domains that may have increasing impact as model intelligence and test-time compute grow.
• Partner with product teams, engineers, partnerships teams, and domain experts to deploy promising methods, measure their effects in real use, and apply findings to further improvements.
• Assess AI’s impacts on human flourishing at the scale of over a billion users.

What This Job Offers

• Job Type: Full-time
• Career Level: Senior
• Education Level: No Education Listed