Research Engineer / Research Scientist - Model Behavior

Model behavior makes current and future models better for people at scale.

This small team of expert tastemakers, designers, PMs, and researchers work across the company and take responsibility for how OpenAI’s models behave. We work on both products with hundreds of millions of users and OpenAI’s internal cutting-edge models.

About the Role

As a model behavior researcher at OpenAI, you will create evals to guide the research organization. The model is the product, and the evals are the model: in this role, you will have singular impact on steering the behavior of today’s products and tomorrow’s AGI.

Model behavior research cross-functionally works with alignment, training, data, RL and post-training. In addition to research, this role demands strong product taste and intuition for classical AI alignment challenges.

In this role, you will:

1. Help decide how today’s products and tomorrow’s AGI will act.

2. Design evaluations to ensure OpenAI’s models are aligned with this intent.

3. Develop new evaluation technology to scalably oversee our models as capabilities advance.

1. Have extensive experience navigating and onboarding to large ML codebases.

2. Have published or worked publicly on language model evals used at big labs.

3. Have artistic, design, or notable product experience in addition to research chops.

4. You understand the gravity of AGI and what it means for us to get its initial conditions right.