Research Engineer, Interpretability

Details

Anthropic’s mission is to create reliable, interpretable, and steerable AI systems to ensure AI is safe and beneficial for users and society. The Interpretability team works to reverse-engineer trained models for a mechanistic understanding, treating neural networks like binary computer programs to 'reverse engineer.'

Responsibilities

• Implement and analyze research experiments on both toy scenarios and large-scale models
• Set up and optimize research workflows for efficiency and reliability at large scale
• Build tools and abstractions to support rapid research experimentation
• Develop and improve tools and infrastructure to support model safety improvements

Requirements

• 5-10+ years of software development experience
• High proficiency in at least one programming language (Python, Rust, Go, Java), with strong productivity in Python
• Experience contributing to empirical AI research projects
• Ability to prioritize impactful work and work comfortably with ambiguity
• Preference for fast-moving collaborative projects over solo efforts
• Interest in machine learning research applications and collaboration with researchers
• Concern for societal impacts and ethics

Strong candidates may also have experience with:

• Designing easily maintainable and bug-free experimental codebases
• Optimizing large-scale distributed systems
• Collaborating closely with researchers
• Language modeling with transformers
• GPUs or PyTorch

Representative Projects

• Building Garcon, a tool for researchers to access LLM internals from Jupyter notebooks
• Optimizing pipelines for collecting and shuffling petabytes of transformer activations
• Profiling and optimizing ML training parallelized across many GPUs
• Facilitating fast launch, manipulation, and analysis of ML experiments
• Creating interactive visualizations of token attention in language models

Benefits and Logistics

• Located in San Francisco office; exceptional remote candidates considered case-by-case
• Bachelor's degree or equivalent experience required
• Hybrid office policy: expected presence in office 25% of the time
• Visa sponsorship available with legal support
• Competitive compensation and benefits, equity donation matching, generous vacation and parental leave, flexible working hours, a collaborative office environment

Company Culture

• Focus on large-scale research efforts with high impact for steerable, trustworthy AI
• Collaborative, empirical approach akin to physics and biology
• Frequent research discussions to maintain focus on highest-impact work
• Strong focus on communication skills and ethical AI development

Salary

$315,000 - $560,000 USD annually