Outbound Communications organization is seeking a highly motivated Senior Research Scientist. In Outbound, we are innovating and building scientific solutions to delight Amazon customers with timely, highly personalized and relevant messages across digital channels such as email, mobile push notifications, WhatsApp and SMS. As a Sr. Applied Scientist in Outbound, you will architect and develop the scientific models for the next generation Outbound message management service that sends 100s of millions of communications (push, email, etc) to worldwide Amazon customers everyday plausibly within minutes of customer-actions on amazon.com. The models handle decisions such as scheduling and optimization across message-ranking, timing and frequency across these messages that drive $20B+ of annual revenue for Amazon. You will use techniques such as reinforcement learning and multi-armed bandits (MAB) to optimize long-term engagement of customers with Amazon. The models and architecture you build ensure a high bar for customer experience through scientifically established guardrails that would use large language models (LLMs). You will own and present research artifacts to internal and external science forums, along with influencing senior leadership and stakeholders for strategic decision-making.
Key job responsibilities
As a Sr. Applied Scientist in Outbound, you will design and develop science models in a highly ambiguous message-management domain that handles scheduling and dispatch decisions of 100s of millions of messages everyday. These models will decide what messages we send to customers, at what times, and through what channels to optimize the long-term customer engagement. You will research and apply state of the art modeling methods such as reinforcement learning, bandits, neural networks, causal modeling methods and LLMs to optimize the message selection, timing and channel of communication for customers. At the crux of architecting these models is influencing the software design of the new message management service responsible to send these messages to customers in span of minutes from when a customer takes actions on-site. You will partner closely with the software engineering and product managers to come up with an ideal design of science architecture that considers design tradeoffs in context of software constraints and user requirements. You will design and analyze experiments that enable quick iteration of model development and learnings as Outbound’s message-management service evolves. You will present scientific findings to senior leaders; and technical, business and product stakeholders to guide for decisions that maximize benefits for Amazon in the long-run. You will mentor junior scientists in the team. You will coordinate with broader Amazon science teams to align your research and scientific design choices with the broader Amazon’s customer objectives. You will document and publish your research-driven artifacts to influence the scientific community in and outside of Amazon.
