10 months ago
Base Salary
$315k - $560k/yr
Responsibilities
- Implement and analyze research experiments in toy scenarios and at large scale in large models.
- Set up and optimize research workflows to run efficiently and reliably at scale.
- Build tools and abstractions for rapid research experimentation.
- Develop and improve tools and infrastructure that enable other teams to use interpretability work for model safety.
- Create and optimize systems for collecting and analyzing transformer activations, profiling ML training, parallelizing workloads across GPUs, and visualizing model attention.
Requirements
- At least 5–10+ years of experience building software.
- High proficiency in at least one programming language such as Python, Rust, Go, or Java, with productivity in Python.
- Experience contributing to empirical AI research projects.
- Ability to prioritize impactful work, operate with ambiguity, and question assumptions.
- Strong collaboration and communication skills and an interest in machine learning research, its applications, and societal impacts.
- A Bachelor's degree in a related field or equivalent experience is required.
- Preferred qualifications include designing experiment-friendly codebases, optimizing large-scale distributed systems, collaborating closely with researchers, language modeling with transformers, and experience with GPUs or PyTorch.
Benefits
- Equity, benefits, and potential incentive compensation are included in the total compensation package.
- Optional equity donation matching, generous vacation and parental leave, flexible working hours, and office space are offered.
- The role is based in the San Francisco office, with exceptional remote work considered case by case.
- Employees are expected to work from an Anthropic office at least 25% of the time under the current hybrid policy.
- Visa sponsorship is available, subject to role and candidate eligibility.
About Anthropic
We're an AI research company that builds reliable, interpretable, and steerable AI systems. Our first product is Claude, an AI assistant for tasks at any scale. Our research interests span multiple areas including natural language, human feedback, scaling laws, reinforcement learning, code generation, and interpretability.