Amazon's Middle Mile Surface Research Science seeks an applied scientist to invent and build machine learning models that improve how Amazon plans and operates its transportation network. Amazon's transportation network moves millions of truckloads of freight between vendors, warehouses, and customers using a fleet of trucks, trains, and airplanes, on time and at low cost. Operating it requires constant decisions about how to route, schedule, and balance capacity across the network, and our strategy is to make those decisions with science-driven technology. Because existing techniques rarely fit Amazon's scale and unique business needs, this role centers on inventing new approaches and algorithms.

As an applied scientist, you'll develop machine learning, forecasting, and prediction models and algorithms. Your first project will focus on trailer imbalance forecasting and safety stock optimization to improve our equipment re-balancing strategy, where you'll own the prediction models and grow your scope over time. Your models will impact business decisions worth billions of dollars and improve the delivery experience for millions of customers.

Key job responsibilities include:

• Design and develop machine learning, forecasting, and other prediction models that enhance our optimization and planning systems.
• Build models and algorithms from prototype to production-level systems.
• Translate ambiguous business problems into modeling approaches, and drive the technical design with product, engineering, and operations partners.
• Influence key business decisions through rigorous modeling and analysis.
• Communicate results and recommendations to scientific and business audiences.

A day in the life includes:

• Analyzing data to investigate a business problem or model performance and identify improvements.
• Brainstorming new algorithmic strategies or business opportunities with fellow scientists.
• Leveraging GenAI to build and test your new model features.
• Running a simulation or experiment to evaluate your model's performance.
• Meeting with product and tech partners to review project requirements, data, design, or other project decisions.
• Reviewing code changes or a design document from a fellow scientist or engineers.
• Writing and presenting a paper documenting algorithm features, results, and recommendations.

About the team: Middle Mile Surface Research Science builds the models and algorithms that plan and operate Amazon's middle mile network. Our work spans operations research, optimization, and machine learning. We work on vehicle route planning, capacity planning, scheduling, network design, transit-time prediction, demand forecasting, and equipment re-balancing. Our team of about ten scientists is part of a broader science organization whose scientists bring deep expertise in machine learning and optimization. We optimize decisions worth billions of dollars and reach millions of customers.