Research Scientist, Interpretability

Job Summary

To advance the understanding of language models, the full-time Research Scientist, Interpretability will develop methods for reverse engineering algorithms, design experiments, and create interpretability features while collaborating with cross-functional teams in a San Francisco office, with potential for remote work for exceptional candidates.

Key responsibilities

• Develop methods for understanding large language models by reverse engineering algorithms learned in their weights
• Design and run robust experiments in both toy scenarios and large models
• Create and analyze new interpretability features and circuits to better understand model operations

Required qualifications

• Bachelor's degree or equivalent in a relevant field
• Strong track record of scientific research, particularly in interpretability
• Experience with Python programming
• Comfort with experimental science and collaborative research environments
• Ability to communicate research findings effectively