Research Engineer / Research Scientist, Finetuning

Anthropic is seeking a Research Engineer / Research Scientist for Finetuning to help construct and rapidly iterate on machine learning experiments to improve the behavior of powerful AI systems. This role involves working on cutting-edge research to make AI helpful, honest, and harmless through techniques like constitutional AI.

Key responsibilities include:

• Developing novel finetuning techniques to improve language model behavior
• Testing constitutional AI techniques at scale and measuring impacts
• Building tooling and infrastructure for efficient fine-tuning experiments
• Developing prompts and strategies to improve and test model behaviors
• Running experiments that feed into key AI research and safety efforts

The ideal candidate will have:

• Significant Python, machine learning, research engineering, or research experience
• A preference for fast-moving collaborative projects with concrete goals
• A results-oriented approach with flexibility and focus on impact
• Willingness to take on tasks outside their job description
• Care about the impact of AI and their work

Strong candidates may also have:

• Prior experience with large language model finetuning (e.g. RLHF)
• Experience with complex codebases and RL infrastructure
• Experience authoring ML/NLP/AI alignment research papers or similar industry work

This role offers the opportunity to do creative, cutting-edge research on frontier models and see concrete improvements in AI performance and safety. The position can be more research or engineering focused depending on the candidate's background.

Anthropic offers competitive compensation including salary, equity, and benefits. They have a hybrid work policy requiring 25% in-office time, with a preference for Bay Area candidates but openness to other locations. The company culture emphasizes collaboration, impact, and advancing the long-term goals of creating steerable and trustworthy AI systems.