Research Scientist, Discovery (Engineering & Applied Sciences)

Anthropic•San Francisco, CA

21h•$500,000 - $850,000•Hybrid

About The Position

Anthropic’s Discovery team aims to build an AI scientist capable of solving long-term reasoning challenges and advancing scientific and technological capabilities. This role focuses on engineering and applied sciences, with the goal of improving Claude’s ability to invent, design, and engineer across disciplines and scales, reliably and safely. The Research Scientist will work across the model stack to develop training environments, rewards, data, and evaluations that improve Claude’s performance on complex engineering and scientific tasks, and bring effective approaches into production training runs. The role involves close collaboration with researchers, engineers, domain experts, and industry partners, including testing model-generated designs and using real-world feedback to shape training and evaluation. Engineering or science experience is a plus but is not required.

Requirements

Proficiency in Python and hands-on experience with a modern machine learning framework such as PyTorch or JAX.
Hands-on experience training or evaluating machine learning models.
A track record of machine learning research, including forming hypotheses, designing and running experiments, and drawing sound conclusions.
Ability to implement research ideas and diagnose problems in complex systems.
Ability to learn unfamiliar technical fields quickly and work productively with experts.
Clear communication skills and a collaborative approach to problem-solving.
Interest in improving Claude’s scientific capabilities and their downstream impact on science and technology.
Commitment to developing AI that is safe and beneficial.
At least a bachelor’s degree or an equivalent combination of education, training, and/or experience.
A field of study relevant to the role, demonstrated through coursework, training, or professional experience.

Nice To Haves

Experience with reinforcement learning and pre-training on large language models, including environment and reward design.
Familiarity with large-scale language model training, evaluation, and inference pipelines.
Published research or practical experience in AI for science or engineering, or in long-horizon reasoning.
A degree or work experience in an engineering or applied science discipline, such as mechanical, electrical, materials, chemical, aerospace, systems, or manufacturing engineering, applied physics, or computational simulation.
Experience building simulators, surrogate models, or simulation-based environments and verifiers, and calibrating and validating them against real-world data.
Experience with large-scale data pipelines and technical or multimodal datasets.
Demonstrated ability to work across multiple domains, such as language modeling, systems engineering, and scientific computing.

Responsibilities

Design, build, and study reinforcement learning environments, rewards, and training data that improve Claude’s capabilities in engineering and applied sciences.
Create benchmarks and evaluation frameworks for realistic engineering and scientific workflows.
Develop simulation-based training environments and verifiers, and work with domain experts to validate them against real-world data.
Develop approaches to long-horizon task completion, multi-agent collaboration, and complex reasoning for engineering and research work.
Research how models can learn efficiently from experiments and real-world feedback.
Study how environments and rewards behave at large training scales, including their robustness, calibration, and generalization, to support capable and well-aligned models.
Scale research ideas from prototype to production training runs.
Collaborate with domain experts and industry partners to build and test Claude’s inventions, designs, and engineering work, and turn feedback into training and evaluation signals.
Help determine where Claude’s engineering capabilities can be used in the real world, what limits adoption, and how this should shape research priorities.