Case Study: EagleView achieves rapid, human-level property damage assessments and reduced losses with Amazon Web Services

Using deep learning on AWS to lower property damage losses from natural disasters

EagleView, a property-data analytics company, tackles the costly delays insurers face after natural disasters—when traditional adjuster surveys can take weeks—by using satellite, aerial, and drone imagery plus deep learning on AWS to produce fast, accurate property-damage assessments for homeowners and carriers.

EagleView captures sub‑1" aerial imagery with a 120+ aircraft fleet, breaks images into geotagged tiles, and runs chained neural networks (built in Apache MXNet) trained on EC2 GPU instances and deployed on ECS, while storing petabytes on S3 and serving results via Redshift/Aurora-backed APIs. The approach yields human-level accuracy (96% per address in a Hurricane Harvey test), enables rapid, actionable responses that can save millions (e.g., targeted tarp deployment after Hurricane Irma), and supports ongoing scaling and framework interoperability with tools like Gluon.

EagleView

Shay Strong

Director of Data Science and Machine Learning