About Anthropic
Anthropic’s mission is to create reliable, interpretable, and steerable AI systems. We want AI to be safe and beneficial for our users and for society as a whole. Our team is a quickly growing group of committed researchers, engineers, policy experts, and business leaders working together to build beneficial AI systems.
About the role
The Knowledge Work team builds the training environments and evaluations that make Claude effective at real-world professional workflows — searching, analyzing, and creating across the tools and documents knowledge workers use every day. As that work scales, the systems behind it need to be as rigorous as the research itself.
As a Research Engineer on Knowledge, you'll design and run experiments that improve how Claude searches, retrieves, and reasons over information at scale. The work spans environment design, data curation, RL training, evaluation, and the infrastructure that supports it all. You'll move fluidly between these depending on what's blocking progress. You'll partner closely with researchers and other RL teams to ship capabilities that show up directly in Claude's behavior.
As our training and evaluations continue to scale, we see a strong synergy between the capabilities our models learn, the tools we build for them to use, and the tools we build for ourselves to understand it all. We own the science behind superhuman epistemics and we ensure the quality of the stack that drives it. We understand that real ownership and impact comes as much through hardening and iterating on environments as it does creating new ones.
Responsibilities
Design, build, and iterate on training environments and data pipelines that improve Claude's ability to reason over knowledge-intensive tasks
Run experiments end-to-end: form a hypothesis, build the infrastructure, train models, analyze results, and decide what to try next
Develop evaluations that meaningfully capture progress on search, retrieval, and reasoning quality
Identify failure modes in current model behavior and translate them into concrete training signals
Collaborate closely with researchers across RL Data, post-training, and product teams to align on priorities and ship improvements
Contribute to shared infrastructure and tooling that compounds the team's velocity over time
Own a clean, canonical set of evaluation tools and processes for Knowledge Work capabilities, including the process used for model releases
Build and automate observability, dashboards, and operational tooling for our training environments and evaluation systems, with an emphasis on high signal-to-noise: a small set of trusted metrics and alerts rather than sprawling instrumentation
You may be a good fit if you
Are a highly experienced Python engineer who ships reliable, well-instrumented code that teammates trust in production
Experience designing, running, and analyzing ML experiments
Ability to work across the stack — from data pipelines to model training to evaluation
Have 5+ years of experience operating ML or distributed systems at scale
Comfort working with ambiguity and choosing the most impactful problem to tackle next
Clear written and verbal communication, especially when collaborating across time zones
Find genuine satisfaction and impact in making existing critical systems dependable
Preferred qualifications
Hands-on experience training, fine-tuning, or doing RL on large language models
Experience building evaluations for LLMs, particularly in open-ended or knowledge-intensive domains
Prior work in a research-heavy environment such as a frontier AI lab, quant research firm, or domain-focused AI startup
Published research on LLMs, RL, retrieval, or related areas
Experience with distributed training systems
Are comfortable being the long-term, context-rich owner of a system and its operational health
Representative projects
Building a training environment that teaches Claude to plan and execute multi-step research tasks against real document co
JobNova