Research Engineer / Research Scientist, Finetuning

About the role:

You want to help construct and rapidly iterate on machine learning experiments to help us improve the behavior of powerful AI systems through finetuning. You care about making AI helpful, honest, and harmless, and are interested in shaping model behavior to be more aligned with human values and goals. You could describe yourself as both a scientist and an engineer. As a Research Scientist or Research Engineer on the Finetuning team, you'll contribute to research on improving language models through techniques like constitutional AI. You will have the opportunity to do creative, cutting-edge research on frontier models, and to see your work result in concrete improvements in performance and safety.

We generally expect research scientists to be able to iterate on their own experiments. We also provide opportunities for engineers to pursue their own research projects. Therefore this role can be more research oriented or more engineering oriented, depending on the experience and interests of the candidate.

Note: Currently, the team has a preference for candidates who are able to be based in the Bay Area. However, we remain open to any candidate who can meet the organization's 25% in-person policy.

Representative projects:

• Help develop novel finetuning techniques to improve language model behavior and make models more helpful, honest, and harmless

• Test out techniques like constitutional AI at scale and measure their impacts on model behavior

• Build tooling and infrastructure to enable efficient fine-tuning experiments on large language models

• Develop novel prompts and prompting strategies to improve and test model behaviors

• Run experiments that feed into key AI research and safety efforts at Anthropic

You may be a good fit if you:

• Have significant Python, machine learning, research engineering, or research experience

• Prefer fast-moving collaborative projects with concrete goals that involve improving model behaviors

• Are results-oriented, with a bias towards flexibility and impact

• Pick up slack, even if it goes outside your job description

• Care about the impact of AI and of your work

Strong candidates may also:

• Haver prior experience with large language model finetuning techniques such as RLHF

• Have experience with complex shared codebases and RL infrastructure

• Have experience authoring research papers in machine learning, NLP, or AI alignment or similar industry experience

Deadline to apply: None. Applications will be reviewed on a rolling basis.