Research Engineer, Interpretability

Overview

Research Engineer (Interpretability) for building infrastructure and tools that enable mechanistic interpretability of neural networks and support production safety audits.

What you'll do

• Build and maintain specialized inference and training infrastructure for interpretability research.
• Resolve scaling and efficiency bottlenecks via profiling and optimization.
• Design tools, abstractions, and platforms to help researchers experiment rapidly.
• Support bringing interpretability research into production safety audits with reliability.
• Work across the stack from model internals to accelerator-level optimization and research tooling.

What you'll need

• Have 5-10+ years of experience building software.
• Be highly proficient in at least one programming language (e.g., Python, Rust, Go, Java) and productive with Python.
• Demonstrate curiosity to learn unfamiliar domains quickly.
• Be able to prioritize impactful work and operate comfortably with ambiguity.
• Collaborate in fast-moving projects and translate research needs into engineering solutions.
• Have interest in interpretability research and its role in AI safety (no formal research experience required).

Details

• Based in the San Francisco office, with case-by-case consideration for remote work.
• Minimum education: Bachelor’s degree or equivalent combination of education, training, and/or experience.