Requirements & tools
Anthropic
The Research Engineer, Knowledge Foundations at Anthropic designs, builds, and iterates on training environments and data pipelines that improve Claude’s ability to reason over knowledge-intensive tasks. Day-to-day responsibilities include running experiments end-to-end (forming hypotheses, building infrastructure, training models, analyzing results, deciding next experiments); developing evaluations that meaningfully capture progress on search, retrieval, and reasoning quality; identifying failure modes in current model behavior and translating them into concrete training signals; collaborating closely with researchers across RL Data, post-training, and product teams; contributing to shared infrastructure and tooling; owning a clean canonical set of evaluation tools and processes for Knowledge Work capabilities; and building automated observability and dashboards for training environments and evaluation systems. Required: highly experienced Python engineer who ships reliable, well-instrumented production code; experience designing, running, and analyzing ML experiments; ability to work across the stack from data pipelines to model training; comfort with high signal-to-noise instrumentation. Bonus: prior experience with RLHF, retrieval-augmented systems, or search infrastructure. Compensation: $350,000 base plus substantial equity (Anthropic stock).
Role context
Research Engineers on the Knowledge Foundations team at frontier AI labs design and run experiments that improve how language models search, retrieve, and reason over information at scale. At Anthropic, this role builds the training environments, evaluations, and infrastructure that make Claude effective at real-world professional knowledge work — searching documents, analyzing data, and creating outputs across the tools knowledge workers use daily. The work spans environment design, data curation, reinforcement learning training, evaluation, and supporting infrastructure. Compensation reflects frontier-AI-lab pay banding at $350,000 base plus significant equity. Hybrid SF office expected. Required: highly experienced Python engineer with experience designing and analyzing ML experiments.
Quick facts
Frequently Asked Questions
How does a Research Engineer differ from a Research Scientist at Anthropic?
Research Scientists focus primarily on the research direction — formulating hypotheses, designing experiments, interpreting results, and shaping the team's intellectual agenda. Research Engineers focus on shipping the infrastructure that makes research happen at scale — training pipelines, evaluation systems, data infrastructure. In practice the boundary is fluid; both roles run experiments and contribute code. Engineers tend to ship more infrastructure; scientists tend to publish more papers. Both roles can grow into Tech Lead positions.
What "Knowledge Foundations" actually means at Anthropic
Knowledge Foundations is the team building training environments and evaluations that make Claude effective at real-world professional workflows — searching documents, analyzing data, creating outputs across the tools knowledge workers use daily. Practically, this means systems for: retrieval-augmented training, multi-document reasoning evaluations, real-world tool-use simulation, and the rigorous evaluation infrastructure to measure capability progress on these dimensions.
Is prior LLM training experience strictly required?
Not strictly. Anthropic specifically mentions that strong Python engineering and experiment design are the harder requirements. Engineers from adjacent fields (search infrastructure, recommendations systems, large-scale ML platforms) can be competitive without specific LLM background. That said, candidates with LLM training experience — even just hands-on fine-tuning — ramp significantly faster on a Knowledge team.