How to apply for Anthropic Fellows Program, ML Systems & Reinforcement Learning

About Anthropic

Anthropic is a frontier AI research and product company working on alignment, policy, and security. The company states that it posts specific opportunities it believes may be high impact, and it links to external concerns about working at a frontier AI lab so candidates can evaluate the tradeoffs.

About the role

The Anthropic Fellows Program provides funding and mentorship to technical talent, regardless of previous experience, to work on an empirical project aligned with Anthropic's research priorities. This ML Systems & Reinforcement Learning track is a 4-month full-time fellowship where fellows primarily use external infrastructure such as open-source models and public APIs, with the goal of producing a public output like a paper submission. In an earlier cohort, over 80% of fellows produced papers.

A typical day

A typical day likely involves running experiments on external infrastructure, reading related work, and meeting with an Anthropic mentor to review progress. The posting does not describe a specific daily routine, so ask current or former fellows about meeting cadence, workspace expectations, and how mentorship is structured.

Who Anthropic is looking for

• Has empirical ML research or engineering experience, especially with reinforcement learning or ML systems, even if not in a formal research role.
• Can design and run experiments using external infrastructure such as open-source models and public APIs, rather than relying on Anthropic's internal compute.
• Is able to work full-time for 4 months and produce a public output, such as a paper submission, by the end of the fellowship.
• Is interested in AI safety and security research and wants to connect with that broader community.

Tips for this application

• State clearly in your application whether the January 2027 start date works for you, since the program says it can sometimes accommodate fellows starting outside the usual cohort timelines.
• Propose an empirical project that uses external infrastructure like open-source models or public APIs and aligns with Anthropic's research priorities in ML systems or reinforcement learning.
• Name the public output you intend to produce, such as a paper submission, because the program measures success by fellows producing public work.
• Apply early. Applications are reviewed on a rolling basis for the next cohort expected to start in January 2027.
• Address the 80000 Hours concerns about working at a frontier AI lab in your own words, since Anthropic links to that material directly on the posting.

What to cover in your cover letter

['A specific empirical project idea in ML systems or reinforcement learning that can be run on external infrastructure.', 'Evidence you can complete a 4-month full-time research project, such as past papers, repos, or experiment logs.', 'Your intended public output, such as a paper submission, and which venue or format you have in mind.', 'Your familiarity with AI safety and security research communities and how you plan to contribute to them.']

Research before applying

• Read Anthropic's mission statement and its stated research priorities in alignment, policy, and security.
• Read the 80000 Hours career review on working at an AI lab, which Anthropic links in the posting.
• Look at papers produced by previous Fellows cohorts to see the expected scope and quality of public output.
• Check what open-source models and public APIs are currently available that could support an ML systems or reinforcement learning project.

Likely interview topics

Based on the job description, expect questions about:

• How you would design an empirical reinforcement learning project using only open-source models and public APIs.
• Your past experience running ML systems experiments, including compute constraints and how you worked around them.
• How you would turn a 4-month project into a paper submission, including scope and timeline.
• Your understanding of Anthropic's research priorities in ML systems and reinforcement learning.
• Your view on the concerns about working at a frontier AI lab, as referenced in the job posting.

Common mistakes to avoid

• Proposing a project that requires Anthropic's internal compute or proprietary models, since fellows primarily use external infrastructure.
• Ignoring the public output requirement. The program expects a paper submission or similar public work, not just internal results.
• Applying without addressing the January 2027 start date or noting if it does not work for you, since the posting asks for this.