Senior Applied Scientist, Amazon

• Master's degree
• Experience programming in Java, C++, Python or related language
• Experience with neural deep learning methods and machine learning
• Experience in building machine learning models for business application
• Experience in applied research

Desirable

• Amazon STEP Science and Tech is seeking Applied (or Research) Scientists. As a key member of the central Research Science Team of logistic operations, these persons will be responsible for designing algorithmic solutions based on data and mathematics for optimizing the end-to-end Amazon supply chain network
• Solve complex optimization and machine learning problems using scalable algorithmic techniques
• Design and develop efficient research prototypes that address real-world problems in the massive logistics network of Amazon
• Lead complex time-bound, long-term as well as ad-hoc analyses to assist decision making
• Communicate to leadership results from business analysis, strategies and tactics
• You will be brainstorming algorithmic approaches with team-mates to solve challenging problems for Amazon logistics operations
• You will be developing and testing prototype solutions with above algorithmic techniques
• You will be scavenging information from the sea of Amazon data to improve these solutions
• You will be meeting with other scientists, engineers, stakeholders and customers to enhance the solutions and get them adopted
• The Science and Tech (SnT) team of EU STEP is looking for candidates who are looking to impact the world with their mathematical and data-driven skills
• We are the End-to-End Supply Chain optimizers. As the core research team, we grow Amazon's logistics business to support decision making in an increasingly complex ecosystem of a data-driven supply chain and e-commerce giant
• Our mathematical algorithms provide confidence in leadership to invest in programs of several hundreds millions euros every year
• Above all, we are having fun solving real-world problems, in real-world speed, while failing & learning along the way
• We use modular algorithmic designs in the domain of combinatorial optimization, solving complicated generalizations of core OR problems with the right level of decomposition, employing parallelization and approximation algorithms
• We use deep learning, bandits, and reinforcement learning to put data into the loop of decision making
• We like to learn new techniques to surprise business stakeholders by making possible what they cannot anticipate. For this reason, we work closely with Amazon scholars and experts from Academic institutions
• We code our prototypes to be production-ready
• We prefer provably optimal solutions than heuristics, though we settle for heuristics when performance dictates it. Overall, we appreciate the value of correct modeling