How to apply for Research Engineer, Knowledge Team

About Anthropic

Anthropic is a frontier AI research and product company working on alignment, policy, and security. Its stated mission is to create reliable, interpretable, and steerable AI systems that are safe and beneficial for users and society. The posting notes that Anthropic does not necessarily recommend working at other positions at the company, and links to external concerns about working at a frontier AI lab.

About the role

This research engineer role sits on the Knowledge Team and focuses on redesigning how Claude interacts with external data sources. The core premise is that existing information architecture paradigms assume human consumers, which no longer holds for LLMs. You will design new architectures from scratch, then train language models via finetuning and reinforcement learning to use those architectures well.

A typical day

The posting does not describe a daily routine. Expect a mix of designing information architecture, writing Python, running finetuning or RL experiments, and building eval sets. Ask the recruiter or hiring manager how the team splits research time versus engineering time, and how the remote setup works in practice.

Who Anthropic is looking for

• Very experienced Python programmer who produces reliable, high-quality code that teammates want to use
• Has solid machine learning research experience, including finetuning and reinforcement learning
• Has built software that utilizes large language models such as Claude
• Can design and evaluate advanced agentic search capabilities and construct hard knowledge base eval sets to expose model failure modes

Tips for this application

• Read the 80,000 Hours career review linked in the posting before applying, since Anthropic explicitly surfaces concerns about doing harm by working at a frontier AI lab.
• Lead with concrete evidence of Python code quality: link to repos, internal tools you built, or libraries teammates adopted.
• Show prior work that uses LLMs in production or research, ideally with Claude or a comparable model. Describe the architecture and what failed.
• Describe any finetuning or RL work with specifics: dataset size, reward design, eval methodology, and measured outcomes.
• Since the role is remote in the US and the team is growing quickly, state your remote-work setup and time zone overlap. Ask the recruiter which time zones the team coordinates across.

What to cover in your cover letter

Argue that current information architectures are built for human readers and explain what you would change for LLM consumers. Give one concrete example of an LLM-based system you built and how you evaluated it. Show finetuning or RL experience with numbers, not adjectives. Explain why you want to work on knowledge and retrieval specifically at Anthropic rather than at another lab.

Research before applying

• Read the 80,000 Hours career review on working at an AI lab, which the posting links directly.
• Read Anthropic's published research on alignment and interpretability to understand the company's technical framing.
• Look at how Claude currently handles external data sources and retrieval, and note where you think the paradigms are wrong.
• Check Anthropic's policy and security work to understand the broader context the Knowledge Team operates in.

Likely interview topics

Based on the job description, expect questions about:

• How would you design an information architecture from scratch for an LLM consumer rather than a human one?
• Walk through a finetuning or RL pipeline you built. How did you define the reward and measure success?
• How would you build a hard knowledge base eval set, and what failure modes would you target?
• Describe an agentic search system you designed or evaluated. Where did it break down?
• How do you write Python that other engineers actually want to use and maintain?

Common mistakes to avoid

• Applying without reading the linked 80,000 Hours career review, given Anthropic explicitly raises concerns about frontier AI lab work.
• Submitting a cover letter that talks about AI safety in general terms without showing Python, ML research, or LLM software experience.
• Claiming finetuning or RL experience without concrete details on datasets, reward design, or eval results.