Research Scientist, Server Demand Forecasting

Meta runs one of the largest server fleets on earth, and every product bet (AI training, ranking, inference, storage) turns into a demand for compute that we must forecast, shape, and match to a physical supply of servers, racks, power, and data center space. As the principal technical leader for Server Demand Planning across Infrastructure, you own the demand side of the plan and the feasible supply requirements it drives: you forecast long- and near-term server capacity demand by rack/hardware type and region, and you build the operations-research models that match that demand to supply so we land the right servers, in the right place, at the right time. You set Meta's technical approach to server demand planning, decide when to ship a production-grade optimization system versus a fast lightweight model to unblock a decision, and set direction across product/service capacity owners, capacity engineering, supply chain, data center planning, and finance, partnering directly with org leaders and growing the technical bench behind you.

Responsibilities

Own the multi-horizon server/MW demand forecast long range (2-5+ years), by resource type and region, and the DC supply requirements it drives Aggregate and normalize demand signals from short term demand and product groups, into a single trusted statistical long term demand plan Formulate and solve the demand-supply matching problem using operations-research models that reconcile forecasted demand with hardware roadmaps, cooling, lead times, and power constraints to inform the Plan of Record Build across the full modeling spectrum: production-grade optimization and forecasting systems and prototype models and heuristics that answer a leadership question or unblock a decision Set functional standards: define the operating model, bridge short- and long-term forecasts, track accuracy and supply-matching metrics, and close forecast-to-actual gaps Partner with data center engineering, site selection, and hardware strategy teams to align next-generation data center designs, rack sizing, and hardware roadmaps with demand, eliminating stranded power, cooling, and space capacity Set the multi-year demand-planning and supply-matching strategy across Infrastructure orgs, spotting capacity risks and opportunities years out and steering decisions before they are expensive to reverse Grow the technical bench: mentor ICs across the planning org and raise the modeling bar company-wide

Qualifications

Bachelor's degree in Computer Science, Computer Engineering, relevant technical field, or equivalent practical experience 8+ years applying operations research / management science to real planning problems in demand planning, capacity planning, supply-demand matching, network optimization, or inventory MS in a quantitative field (Operations Research, Industrial Engineering, Applied Math, Statistics, CS, or related), or equivalent experience Operations-research toolkit including: mathematical optimization (LP, MILP, stochastic/robust optimization), simulation, queuing theory, and probabilistic/statistical forecasting Demonstrated ability to build BOTH production-grade models/systems (deployed, maintained, driving real decisions) AND lightweight/prototype models delivered fast under ambiguity Experience reconciling forecasted demand against constrained supply, lead times, and inventory in a planning or capacity environment Fluency with optimization solvers (Gurobi, CPLEX, Xpress, or OR-Tools) and with SQL + Python for modeling, analysis, and pipelines Direct experience with server/compute or data center capacity planning at hyperscale Experience with planning platforms (Kinaxis, SAP IBP, o9, Blue Yonder, Demantra, E2open) and internal capacity tools (internal capacity planning and supply-matching tools) Statistical/ML forecasting depth (time series, hierarchical/probabilistic forecasting, forecast reconciliation) PhD in Operations Research, Industrial Engineering, Management Science, or a related quantitative field Publications, patents, or recognized technical leadership in OR / optimization / forecasting