Anthropic Fellows Program, ML Systems & Performance

The role

This fellowship sits within Anthropic's broader fellows initiative and focuses on the systems side of machine learning. Participants take on an empirical project, mostly using outside infrastructure, and are paired with mentors from the research team. The role blends engineering craft with research exploration.

What you would do

• Construct simulators that model hardware accelerator behavior for workload analysis
• Extend open-source projects by adding support for additional accelerator hardware
• Develop on-demand infrastructure that other infrastructure-heavy fellow projects can draw on
• Assemble large synthetic data and environment pipelines for experiments
• Work through a project selection and mentor pairing process at the start

What they asked for

• Fluent in writing Python
• Able to commit to full-time participation for the entire program
• Hold work authorization in the UK, US, or Canada and reside there during the term
• Background in computer science, mathematics, or physics
• Comfortable implementing ideas rapidly and explaining them clearly

Nice to have

• Solid software engineering background including complex ML system builds
• Skill in balancing exploratory research with engineering discipline and reliability
• Experience with distributed or high-performance computing, such as in trading environments
• Hands-on work with training, adjusting, or assessing large language models
• Competence diagnosing problems in model training runs