Staff+ Research Engineer (RL Data Platform)

• Anthropic’s RL Data Platform team builds the systems that produce, move, and serve the human data Claude learns from: the interfaces humans use to give feedback, the pipelines that turn raw feedback into training signal, and the tooling researchers use to launch, monitor, and inspect data collection

• Every RL run depends on a steady supply of high-quality data - human feedback, expert demonstrations, graded transcripts - and when a researcher has an idea for new data on Monday, our job is to make it collectable by Wednesday and in the training mix by Friday

• This is a full-stack, ownership-heavy role on a small, senior team

• You’ll design and ship web interfaces used by thousands of expert annotators, build the backend services and data pipelines behind them, and work directly with RL researchers to understand what data they need and why. You’ll scope your own projects, make architectural calls, and see them through to production

• We’re looking for engineers who treat researchers as their users, build for reliability first, and care as much about the shape of the data leaving the system as the UI going into it

• Design, build, and operate the feedback and data collection interfaces used by human annotators, domain experts, and internal researchers

• Build and maintain the backend services, APIs, and pipelines that route model samples to humans and return structured feedback to training

• Own the reliability, latency, and usability of systems that run continuously against live model endpoints

• Partner with RL researchers to translate loosely specified data needs into well-scoped collection campaigns and the tooling to run them

• Build dashboards, monitoring, and inspection tools so researchers can see data quality and throughput without asking an engineer

• Identify and remove the bottlenecks between “we want this data” and “it’s in the training mix”

• Representative projects:

• Build an interface that lets a domain expert review a long agentic transcript, flag the step where things went wrong, and write a corrected continuation - with the result landing in a training-ready format

• Rework the sampling path between our feedback interfaces and model endpoints to cut time-to-first-sample for annotators

• Build a campaign launcher that lets a researcher stand up a new data collection effort (task, rubric, population, quality checks) without writing code

• Instrument annotator behaviour to detect low-effort or adversarial work and surface it to the quality team automatically

• Design the data model for a kind of feedback we haven’t collected before, and ship the pipeline that gets it into the training mix

Benefits

• Comprehensive health, dental, and vision insurance for you and your dependents

• Inclusive fertility benefits via Carrot Fertility

• 22 weeks of paid parental leave

• Flexible paid time off and absence policies

• Mental health support for you and your dependents

• Competitive salary and equity packages

• Optional equity donation matching at a 1:1 ratio, up to 25% of your equity grant

• Retirement plans with competitive matching

• Life and income protection plans

• $500/month flexible wellness and time saver stipend

• Commuter benefits

• Annual education stipend

• Home office stipends

• Relocation support for those moving for Anthropic

• Daily meals and snacks in the office- Care about the societal impacts of your work

• Comfort working directly with technical stakeholders whose needs change week to week, and the judgment to push back when something isn’t worth building

• Strong full-stack engineering skills, with production experience in TypeScript/React on the frontend and Python on the backend

• A track record of owning projects end-to-end, from an ambiguous brief to something in production that people use

• Effective use of AI tools in your own day-to-day work

• Experience designing and operating backend services and data pipelines that other teams depend on

• We encourage you to apply even if you do not believe you meet every single qualification. Not all strong candidates will meet every single qualification as listed

• Experience with RLHF, preference data, or other human-feedback pipelines for ML systems

• Experience building annotation, labelling, evaluation, or other human-in-the-loop data tooling

• Experience shipping researcher-facing or other expert-facing internal tools people love: interviewing users, hunting down friction, measurably improving the experience

• Experience running experiments on data collection interfaces and using the results to improve data quality

• Experience working with crowdworker or expert vendor platforms at scale

• Familiarity with how LLMs are trained and evaluated