Are you fascinated with the idea of creating a digital representation of the world? What about building learning systems to create the most precise and accurate representation? Join the Geospatial Data team to build Maps for Amazon.

We build learning systems to identify road networks, POIs, and geocodes of addresses worldwide. Our delivery operations use these systems to determine the locations and plan the routes for delivery. Drivers use these systems to navigate to the delivery locations. Our mission is to vend geospatial data (e.g. maps, traffic, addresses, and locations/geocodes) that is both authoritative and fresh through an intuitive experience that enables every driver – independent of tenure and affinity – to succeed in their delivery tasks.

We are not creating another consumer-grade mapping solution, we are building systems that enable depth focused solutions. For example, we are interested in not only getting a person to an address like 300 Boren Ave N, we are also interested in helping them park, optimally group the packages for that stop based on proximity, find out if there is a mailing room in the building that is open at that time, and help them navigate quickly to that mailing room or alternative location. We are also interested in accurately estimating how long it would take. We incorporate the ability to leverage multiple modes of transportation and traffic awareness to find the most efficient paths for our drivers. We are also interested in making it easy to calculate paths to cover hundreds of delivery points. Several of these problems require us to build systems that can work with an ensemble of models as well as support the right segmentation of inputs to make good estimates on the outputs.

There are several unsolved or partially solved problems in this space; such as automatically adding new roads detected from sensor/video data into the larger road graph, detecting if a new road is in fact just a modification to an existing road (such as a change in curvature of an existing road due to a new sidewalk), accurately determining the bearing of a person when they start traveling by leveraging IMU sensor source, parsing unstructured addresses such as in countries like India, processing alternate solutions within microseconds on a mobile device without talking to a backend service and so on.

The technical domain is multi-faceted. We build low-latency, highly-available services, we build big data processing pipelines to refresh or produce new data, we train ML models, and we run science experiments. We also build platforms for these capabilities, ranging from ML and experimentation platforms to semantic graph data stores that links billions of geospatial artifacts for use by a number of different use cases.

Our key output metrics include location accuracy, coverage and accuracy of our road network for routing users to the correct location, predictive accuracy of service, and transit estimates. We also measure the operational impact of these inputs on delivery success and on the gaps between actual versus planned on zone times, transit times, and service times.

If you have an entrepreneurial spirit, know how to deliver, are deeply technical, highly innovative and long for the opportunity to build pioneering solutions to challenging problems, we want to talk to you.

Key job responsibilities
Participate in the design, implementation, and deployment of successful large-scale systems and services in support of our fulfillment operations and the businesses they support.

Participate in the definition of secure, scalable, and low-latency services and efficient physical processes.

Work in expert cross-functional teams delivering on demanding projects.

Functionally decompose complex problems into simple, straight-forward solutions.

Understand system inter-dependencies and limitations.

Share knowledge in performance, scalability, enterprise system architecture, and engineering best practices.

A day in the life
Collaborate with engineering and science teams on building state-of-the-art scalable, big data pipelines, ML Platform (for model training, lifecycle management and inference), high-availability services, and semantic graphs with geospatial data.

Analyze nuanced metrics to understand system behavior and innovate on solutions to optimize our success metrics.

Play a key role in the technology that empowers the delivery of millions of smiles every day around the world.

About the team
This role is within the LastMile, Geospatial Data organization. Geospatial data (e.g., maps, traffic, addresses, and locations) are foundational inputs enabling us to plan safe, efficient routes and guide drivers to their delivery/pickup points. The team for this position is developing a knowledge graph, pulling together foundational geospatial data sets including addresses, geocodes, and maps data from a shared ontology across LastMile Teams. Linking these elements enables our organization to achieve the vision to vend geospatial data that is authoritative, complete, and fresh through an intuitive experience that enables every driver to succeed in their delivery tasks.