Data Scientist, Finance Forecasting

About the role

As an early member of our Finance Analytics and Business Intelligence team, you will play an instrumental role in our company's mission of building safe and beneficial artificial intelligence by establishing robust analytics engineering and business intelligence capabilities for our Finance & Accounting organization. In this unique company, technology, and moment in history, your work will be critical to ensuring accurate financial reporting, streamlining accounting processes, and supporting our financial operations as we deploy safe, frontier AI at scale to the world.

As a founding Data Scientist on Anthropic's Finance Forecasting team, you'll own revenue forecasting models that drive capacity planning and board reporting, build the backtesting and accuracy discipline that makes those forecasts trustworthy, and lead causal measurement work that tells us how much a given launch or event actually moved revenue — turning what is currently a best-guess conversation into a designed, repeatable readout.

We're hiring for two seats on this team, and we expect candidates to lead with depth in either production time-series forecasting or causal inference. Both shapes are first-class; tell us in your application which one feels more like you.

This is a build role on a small team. The forecast is visible inside the company, and when it's wrong, people notice. We're looking for someone who would rather have their models scored than stay vague, and who is excited to set the bar for how forecasting is done here.

Responsibilities:

• Own a core piece of the team's modeling work — either the production revenue forecasts themselves (scoping, development, backtesting, deployment, monitoring) or the causal measurement program for launches and events
• Build and run backtesting and accuracy tracking for your models, and use the results to improve quality cycle over cycle
• Contribute to the team's broader research direction, including event-aware forecast architectures, hierarchical reconciliation, and causal designs that hold up under launch-driven step-changes
• Translate model output and accuracy results into clear recommendations for Finance and executive leadership
• Partner with the team's Analytics Engineer on feature pipelines, model deployment, and the forecast store

You may be a good fit if you:

• Have substantial experience in data science, forecasting, or quantitative finance, including time owning models in production rather than only in notebooks
• Are deeply fluent in Python and SQL and comfortable productionizing what you build
• Have a strong applied statistics foundation, with depth in either production time-series methods (Prophet, ETS, ARIMA, gradient-boosted approaches, neural forecasting, hierarchical reconciliation) or causal inference (difference-in-differences, synthetic control, Bayesian structural time series, event studies)
• Have built backtesting and accuracy-tracking discipline before and are comfortable having your models scored publicly
• Have presented and defended a forecast or causal estimate to executives
• Have a bias for action and do your best work in ambiguous, early-stage environments

Strong candidates may also have:

• Experience with exogenous-regressor or event-aware forecasting approaches
• Familiarity with hybrid or foundation-model forecasting (TimeGPT-class systems)
• Background in pricing and elasticity modeling or marketing-mix modeling
• Experience forecasting a consumption-based or usage-billed business (cloud, API, marketplace)

What success looks like in your first year:

• A cross-methodology accuracy baseline is published as a standing Finance metric
• The base revenue forecast is measurably more accurate than today's consensus on the same out-of-sample window
• Launch incrementality is a standing readout with a designed, backtested method behind it
• Forecasts are queryable from the forecast store and feeding capacity planning and board reporting

Things to know

Anthropic's business is growing quickly, with launch-driven step-changes that mean what worked last quarter may not generalize. Methodology direction is a research program, not a settled architecture. You'll be building practices and systems while also delivering a monthly number, alongside an Analytics Engineer who is building the platform underneath you.