Applied Scientist , Inbound Systems

Amazon; New York (NY); On-site; USD 150,000 - 210,000; Full time

Job summary

Amazon's Supply Chain Planning Optimization team is building the next generation of planning systems at planetary scale, solving large-scale optimization with hundreds of thousands of variables in seconds, aligned with probabilistic forecasts for real inventory movement across hundreds of millions of customers.

We seek an experienced scientist with depth in large-scale optimization, stochastic methods, and forecasting to own end-to-end models from whiteboard to production in a cloud-native

Qualifications

• Experience in large-scale optimization, with stochastic methods and decomposition.
• Fluency in machine learning and probabilistic forecasting.
• Track record delivering complex scientific systems end to end.

Responsibilities

• Build state-of-the-art, scalable stochastic optimization and probabilistic forecasting algorithms.
• Shape formulation, decomposition, and solution for computational performance at scale.
• Engineer production-grade, cloud-native software with modern development practices.
• Architect long-term scientific solutions while shipping incremental improvements.
• Prototype quickly, drive pilots, and iterate with operational feedback.
• Deliver science into production by partnering with internal customers and leadership.

Skills

Large-scale optimization

Stochastic optimization

Decomposition methods

Machine learning

Probabilistic forecasting

End-to-end delivery

Job description

Description

Amazon's Supply Chain is the backbone of the fastest growing e-commerce business in the world, and planning it is one of the largest optimization problems in industry. Every week we decide where millions of products should sit across hundreds of fulfillment centers, how inventory should flow between suppliers, buildings, and customers, and how to balance cost, speed, and capacity, all under deeply uncertain demand. No single model can solve a problem this large. The Supply Chain Planning Optimization team is building the next generation of planning systems around large-scale distributed optimization: decomposing the full network problem into tractable pieces that coordinate toward a globally consistent plan, solving optimization problems with hundreds of thousands of variables in seconds, and pairing them with probabilistic forecasts so plans hold up when reality diverges from the forecast. The work spans the full stack of modern operations research, from decomposition and convergence to stochastic optimization and solver performance at scale. And it has a rare property: the models you build move real inventory for hundreds of millions of customers, and you see the results in the physical world within weeks.

What you\'ll do

• Design and deploy large-scale optimization and forecasting models that plan inventory placement and flow across our EU/NA fulfillment network under uncertainty
• Shape how the full network problem is decomposed and coordinated, defining the mathematical architecture of the planning system rather than just the models within it
• Push the computational frontier through formulations that solve fast and reliably at scale, and through the solver technology and tooling that make experimentation cheap
• Work with science, engineering, operations, and finance partners to take ideas from whiteboard to production, then own them end to end once live

What we\'re looking for

An experienced scientist with depth in large-scale optimization (stochastic optimization and decomposition methods especially welcome), fluency in machine learning and probabilistic forecasting, and a track record of delivering complex scientific systems end to end. You care about both the elegance of a formulation and whether it solves in two seconds or two hundred, and you\'re energized by delivering incremental wins while building toward a long-term scientific vision.

If you want your optimization theory to move real inventory at planetary scale, this is the team.

Key job responsibilities

• Build state-of-the-art, robust, and scalable stochastic optimization and probabilistic forecasting algorithms that drive optimal planning and execution under uncertainty across Amazon\'s end-to-end supply chain
• Shape how large-scale planning problems are formulated, decomposed, and solved — designing for computational performance and reliability at the scale of Amazon\'s fulfillment network
• Engineer your algorithms as production-grade, cloud-native software, applying modern development practices from prototype through deployment
• Think several steps ahead: architect long-term scientific solutions while continuously shipping incremental improvements to what\'s already running
• Prototype fast, drive early adoption through pilots, integrate operational feedback, and iterate
• Deliver your science into production by partnering closely with internal customers — understanding their needs and blockers, and influencing their roadmaps
• Lead complex analyses and communicate results and recommendations crisply to senior leadership
• Stay at the frontier as an active member of the science community: research, apply, and publish (internally and externally) the latest OR/ML techniques from academia and industry

About the team

We are a team of scientists and engineers who believe that some of the hardest optimization problems in the world are hiding inside everyday questions like "where should this product sit so a customer gets it tomorrow?" We take ideas from the frontier of operations research and machine learning — distributed optimization, planning under uncertainty, solving at massive scale — and turn them into systems that steer one of the largest supply chains on Earth. If a model we ship on Monday moves millions of units by Friday, that\'s a normal week; that loop between theory and the physical world is why we\'re here.