Research Engineer, Human-Centered AI

OpenAI is dedicated to ensuring that our AI systems are safe, trustworthy, useful, and consistently aligned with human values, even as they scale in complexity and capability. Our research focuses on methods that help AI reliably follow human intent in diverse scenarios—including adversarial or high-stakes settings—and ultimately maximize its benefit to individuals and society.

A critical component of our approach involves leveraging high-quality human and synthetic data to train and evaluate models, ensuring alignment techniques remain effective as capabilities grow. We work with experts across domains and modalities to develop scalable methods for generating high-quality data, and using it to produce safe and useful models.

The two pillars of our approach are: (1) harnessing improved capabilities into alignment, making sure that our alignment techniques improve, rather than break, as capabilities grow, and (2) centering humans by developing mechanisms and interfaces that enable humans to both express their intent and to effectively supervise and control AIs, even in highly complex situations.

We're looking for individuals with strong ML engineering skills, research experience, and a deep understanding of Human-Machine Interaction challenges, especially with novel and highly capable models.

About the Role

As a Research Engineer, you will:

• Research and model mechanisms that create value for people, with an emphasis on explaining or predicting preferences, behaviors, and satisfaction.
• Quantify the nuances of human behavior and capture them in data-driven systems, whether by designing advanced labeling tasks or analyzing user feedback patterns
• Design robust evaluations for measuring alignment and real-world utility, iterating quickly to uncover what makes certain feedback and training protocols more effective.
• Design and evaluate new Human-AI-interaction paradigms and scalable oversight methods that redefine how humans interact with, understand, and supervise our models.
• Develop and evaluate alignment capabilities that are subjective, context-dependent, and hard to measure.

You might thrive in this role if you:

• Have experience with machine learning frameworks (e.g., PyTorch) and are comfortable experimenting with large-scale models.
• Enjoy moving fluidly between high-level research questions and low-level implementation details, adapting methods to solve ambiguous, dynamic problems.
• Are goal-oriented instead of method-oriented, and are not afraid of tedious but high-value work when needed.
• Have an interest or background in cognitive science, computational linguistics, human-computer interaction, or social sciences.
• Are strongly motivated by OpenAI's mission of building safe, universally beneficial AGI and are aligned with OpenAI's charter
• Want to work on systems that balance breakthrough capabilities with robust alignment, ultimately shaping a safer and more human-centered AI landscape.
• Excel in fast-paced, collaborative, and cutting-edge research environments.