About the role
About Anthropic
Anthropic’s mission is to create reliable, interpretable, and steerable AI systems. We want AI to be safe and beneficial for our users and for society as a whole. Our team is a quickly growing group of committed researchers, engineers, policy experts, and business leaders working together to build beneficial AI systems.
About the role
The Discovery team at Anthropic is organized around the north star goal of building an AI scientist – a system capable of solving the long-term reasoning challenges and basic capabilities necessary to push the scientific and technological frontier.
We're expanding our work on engineering and the applied sciences. Our goal is to contribute to training Claude to invent, design, and engineer across every discipline and scale, reliably and safely: new manufacturing and experimental techniques, processes, materials, devices, instruments, machines, energy systems, and factories. We want this work to accelerate science and technology and benefit people widely.
To achieve these goals, we're seeking exceptional machine learning researchers with deep expertise in training and evaluating frontier models. In this role, you'll work end to end across the model stack, developing training environments, rewards, data, and evaluations that measure and improve Claude's performance on complex engineering and scientific tasks, and carrying what works into production training runs. You'll collaborate closely with researchers, engineers, and partners across Anthropic and industry to build AI systems that can engage in all phases of research and development, and that learn from real-world feedback. A background in engineering or the sciences is a plus but not required.
Key responsibilities
- Design, build, and study reinforcement learning environments, rewards, and training data that improve Claude's capabilities in engineering and the applied sciences
- Create benchmarks and evaluation frameworks that measure model performance on realistic engineering and scientific workflows
- Develop simulation-based training environments and verifiers, working with domain experts to validate them against real-world data
- Develop approaches to long-horizon task completion, multi-agent collaboration, and complex reasoning essential for real engineering and research work
- Research how models can learn efficiently from experiments and real-world feedback
- Study how environments and rewards behave in large-scale training, including robustness, calibration, and generalization, so that training produces models that are both capable and well aligned