Staff Machine Learning Engineer, Ads Forecasting & Scenarios

The opportunity

Build the machine learning capabilities that help advertisers understand what they could achieve on Reddit and make better investment decisions.

You will lead models that estimate how campaign outcomes could change with different budgets, bids, audiences, creative performance, or objectives. The work connects supply and demand forecasting, audience availability, response prediction, and marketplace simulation to help advertisers compare alternatives and recognize valuable opportunities they may be missing.

This is a hands-on technical leadership role with ownership of modeling direction, evaluation standards, and reusable production capabilities across advertiser planning and recommendations.

What you’ll do

• Lead forecasting and scenario modeling. Translate advertiser questions into models of delivery, reach, performance, and investment tradeoffs. Set technical direction and guide execution across the team.
• Connect inventory, demand, and response. Combine forecasts of available advertising opportunities, audience activity, marketplace competition, and predicted response to estimate achievable delivery and performance.
• Estimate outcomes under alternative choices. Develop models that account for changes in budget, bid, audience, and optimization objective, including interactions among these choices, auction dynamics, and campaign constraints.
• Quantify unrealized opportunity. Incorporate creative-performance estimates from partner teams to assess how stronger creative could have improved campaign results. Translate those inputs into potential gains in delivery, efficiency, or advertiser value.
• Build models that generalize. Apply probabilistic forecasting, multitask learning, transfer learning, hierarchical modeling, or related approaches to handle sparse segments, cold start, delayed feedback, and changing conditions.
• Establish trustworthy comparisons. Evaluate numerical accuracy, calibration, uncertainty, and cohort-level performance. Define supported scenarios and identify when additional evidence or experimentation is needed.
• Close the learning loop. Compare recorded predictions with mature outcomes, diagnose systematic errors, and validate scenario estimates through experiments and observed results.
• Own production delivery and technical leadership. Partner with engineers on reliable training, serving, monitoring, and iteration. Lead designs, mentor engineers, and collaborate with Product, Data Science, Ads Delivery, Audience, and Creative teams.

What we’re looking for

• A track record of developing and shipping predictive ML systems in advertising, marketplaces, recommendations, search, demand forecasting, or similarly complex domains.
• Strong foundations in statistical learning, probability, and optimization, with experience modeling large-scale behavioral or transactional data.
• Depth in one or more relevant areas: probabilistic or hierarchical forecasting, calibrated conversion/value prediction, multitask or transfer learning, delayed-feedback or survival modeling, or response modeling.
• Strong judgment in evaluation, including temporal validation, calibration, segment-level performance, and the assumptions required to estimate outcomes under changed conditions.
• Experience addressing sparse observations, cold start, seasonality, distribution shift, selection bias, or partially observed labels.
• Strong programming and software engineering skills, including experience building and operating production ML systems.
• Demonstrated ability to lead ambiguous technical initiatives, make practical tradeoffs, and develop other engineers.

Preferred experience

Complementary depth in one or more of:

• Demand, supply, and inventory forecasting: Advertising inventory, marketplace demand, traffic, retail demand, or capacity forecasting. Relevant methods include probabilistic time-series models, hierarchical/global forecasting, external-signal modeling, and multi-horizon prediction.
• Auction and investment modeling: Auction simulation or replay, conditional win-rate and price models, bid/budget response curves, or constrained optimization.
• Scenario evaluation: Counterfactual or off-policy evaluation, sensitivity analysis, uncertainty propagation, or composition of multiple predictive models.
• Reach and audience estimation: Population and overlap estimation, probabilistic reach/frequency models, sampling correction, or saturation modeling.
• Advanced predictive methods: Mixture-of-experts, distributional or quantile prediction, temporal sequence models, or state-space models.

What success looks like

Accurate and calibrated forecasts across priority advertiser populations; credible comparisons of supported campaign alternatives; reusable capabilities across planning and recommendations; and evidence that advertisers identify valuable opportunities and make better investment decisions.

#LI-AK1