How to apply for Research Engineer, Machine Learning (RL Velocity)

About Anthropic

Anthropic is a frontier AI research and product company focused on building reliable, interpretable, and steerable AI systems. Teams work across alignment, policy, and security, and the company posts specific high-impact opportunities rather than recommending any role. It is a place for people who want to work directly on making advanced AI systems safer and more beneficial.

About the role

The RL Velocity team owns the efficiency and reliability of Anthropic's RL Science stack, which includes the infrastructure, tooling, and systems researchers use to iterate on training runs. As a research engineer, you will build and improve the core platform, remove bottlenecks, and own the reliability and performance of research runs end-to-end. Small improvements to velocity compound across every researcher and every run, so the work directly shapes how fast Anthropic can ship better models.

A typical day

A typical day might involve profiling an RL training run to find a slowdown, pairing with researchers to understand where their workflow is stuck, and shipping a tooling change that reduces iteration time. You would also coordinate with adjacent teams like inference or sandboxing to fix cross-stack issues and review design decisions for how RL is done at scale. The exact routine is not described in the posting, so ask the hiring team how the RL Velocity team structures its week and how it prioritizes reliability versus new tooling.

Who Anthropic is looking for

• Strong software engineering skills with experience building or maintaining ML training infrastructure.
• Comfortable debugging, profiling, and rearchitecting systems to remove bottlenecks across a stack.
• Able to partner closely with researchers and adjacent engineering teams such as inference and sandboxing to understand pain points and ship tooling.
• Has owned reliability and performance of research or production systems end-to-end, and can contribute to design decisions for RL at scale.

Tips for this application

• Read Anthropic's career review on 80,000 Hours before applying, since the company explicitly links to concerns about working at a frontier AI lab and expects candidates to have considered them.
• Tailor your resume to show work on RL infrastructure, tooling, or systems that made researchers or training runs faster, not just general ML modeling work.
• Use the cover letter to name specific bottlenecks you have removed in past training stacks, such as debugging, profiling, or rearchitecting, because the role is defined by velocity and reliability.
• If you have worked with inference, sandboxing, or adjacent platform teams, describe how you partnered with them, since the job description calls out those collaborations directly.
• Check Anthropic's published research and engineering blog posts on RL and infrastructure so you can speak to how your experience maps to their stack in interviews.

What to cover in your cover letter

Explain why you want to work on RL infrastructure efficiency rather than only on model research. Give one or two concrete examples of bottlenecks you removed and how that improved iteration speed for a research team. Show that you understand the RL Velocity team owns reliability and performance end-to-end, and describe how you have owned systems in that way before. Mention how you have worked with adjacent teams like inference or sandboxing to ship tooling that researchers actually use.

Research before applying

• Read Anthropic's mission statement and its published work on alignment, interpretability, and steerability to understand the context for RL infrastructure work.
• Review the 80,000 Hours career review on working at an AI lab, which Anthropic links directly in the job posting.
• Look at Anthropic's engineering or research blog for any posts on RL training, infrastructure, or velocity-related tooling.
• Find public talks or papers from Anthropic researchers on RL methods to understand what the RL Science stack supports.

Likely interview topics

Based on the job description, expect questions about:

• How you would profile and remove a bottleneck in an RL training stack, with a concrete example from your past work.
• Your experience building or maintaining infrastructure that researchers depend on daily, and how you measured its impact on iteration speed.
• How you have partnered with researchers or adjacent engineering teams to understand pain points and ship tooling.
• A time you owned reliability and performance of a research or production system end-to-end, including what broke and how you fixed it.
• Your view on design decisions that shape how RL is done at scale, and how you would approach tradeoffs between velocity and reliability.

Common mistakes to avoid

• Applying without addressing the concerns Anthropic raises about working at a frontier AI company, since the posting explicitly links to that material.
• Focusing only on model research or algorithm design and ignoring the infrastructure, tooling, and reliability aspects that define this role.
• Treating the role as generic ML engineering and failing to show how your work made researchers or training runs faster.