Lead Data Scientist, Demand

Farmer's Fridge Careers•Chicago, IL

13d•Onsite

About The Position

Farmer’s Fridge is hiring its first full-time data scientist dedicated to demand planning. The Lead Data Scientist, Demand owns the predictive layer of the company’s planning process: forecasting what will sell, where, and when. The role maintains and improves existing fridge and retail forecasting models in production, rebuilds them when a better approach is justified, and expands into company-level volume and revenue forecasting as the core models mature. It works closely with the Lead Decision Scientist, Data Engineering, the demand planning team, and the Manager, Demand Planning. The role is based in Chicago or requires relocation to Chicago and is in-office. The position forecasts item-level demand by location and day across more than 2,000 fridges and retail partner stores. Models must account for low-volume and intermittent demand, stockouts that obscure true demand, short shelf life, new locations and products, holidays, and promotions. Forecasts should be probabilistic and support decisions that balance lost sales against waste. Model outcomes are evaluated through business measures such as sell-through, unsolds, stockouts, and contribution margin—not statistical accuracy alone. The Lead Data Scientist partners with Channel Strategy and Demand Planning, which set business rules and commercial objectives, and with the Lead Decision Scientist, whose optimization models use forecasts to make stocking, allocation, and routing decisions. The role also works with Data Engineering to operate models in production pipelines, and with demand planning, supply chain, and operations to explain forecast expectations, uncertainty, and tradeoffs. It sets standards for forecasting model development, testing, documentation, and deployment.

Requirements

4–8 years of experience building and maintaining forecasting or predictive models in production.
Deep demand forecasting experience using time-series, hierarchical, or machine-learning methods such as gradient boosting, state space, or Bayesian methods.
Comfort with count data and intermittent demand, such as Poisson or negative binomial methods.
Experience correcting for demand censored by stockouts, or a strong understanding of why this matters.
Strong Python and SQL skills.
Experience with Snowflake or a similar cloud data warehouse.
Rigor in backtesting and experiment design, with the ability to demonstrate that a model change helped.
Track record of owning model work end to end, from problem framing through production and monitoring.
Ability to identify a model change shipped and the business result it produced in dollars.
Clear communication skills, including explaining forecast uncertainty and tradeoffs to planners and operators and defending modeling choices to technical peers.
Comfort working in a fast-paced environment and setting direction.
Familiarity with using AI tools in day-to-day work.
Based in Chicago or willing to relocate; this is an in-office role.

Nice To Haves

Forecasting experience with perishable goods, grocery, CPG, food service, or multi-location retail.
Exposure to forecasts feeding optimization models for inventory, allocation, or routing.
Causal inference experience, such as difference-in-differences or matched controls, for measuring pricing, promotion, or operational changes.

Responsibilities

Own fridge and retail demand forecasts end to end, including data, features, modeling, validation, and production.
Build forecasts that maximize contribution margin dollars and track impact through sell-through, unsolds, and stockouts against Channel targets.
Monitor systematic forecast bias by item and fridge and provide optimization models with realistic ranges for achieving business goals.
Maintain and improve existing production models, rebuilding them where a better approach clearly pays off.
Forecast at the item, location, and day level across more than 2,000 fridges and retail partner stores, including low-volume and intermittent demand.
Model stockouts and censored demand, short shelf life, new fridge ramp-up, new menu items, holidays, and promotions.
Produce probabilistic forecasts for optimization models to weigh lost-sale costs against waste costs.
Partner with the Lead Decision Scientist on forecast outputs required by the solver and measure how forecast changes affect stocking decisions and outcomes.
Build backtesting and monitoring; evaluate models using business outcomes as well as statistical error.
Maintain a model scorecard and review it on a set cadence with Channel Strategy and Demand Planning.
Measure every material model change before and after on sell-through, unsolds, and stockouts, translating results into contribution margin per fridge per week.
Partner with Data Engineering to run models in production pipelines using Snowflake, dbt, and Dagster without manual babysitting.
Work with demand planning, supply chain, and operations to explain model expectations, reasoning, and uncertainty.
Measure the value of planner overrides, incorporate consistently helpful overrides as explicit rules, and retire overrides that do not help.
Expand into company-level volume and revenue forecasting as core models mature.
Maintain a living record of model business rules, including what each rule does, its origin, owner, and last change.
Set standards for how forecasting models are built, tested, documented, and shipped at Farmer’s Fridge.