Tripadvisor built a social distancing application called Crowdfree that enables users to see people’s presence in stores and public spaces. The app leverages serverless architecture, purpose-built databases, and data lakes. AWS Cloud Amazon Elasticsearch Service Lambda function POI data from Safegraph S3 bucket Telemetry data from Veraset S3 bucket 4 Telemetry data from X-mode S3 bucket Amazon EMR EMR Opensource geographical data Lambda function Amazon DynamoDB 8 Amazon EC2 3 Lambda function EMR 10 Amazon API Gateway 7 5 S3 bucket AWS WAF 6 Amazon Athena 2 Amazon RDS (PostgreSQL) Amazon Aurora (PostgreSQL) 9 AWS Amplify AWS Marketplace (OpenStreetMaps) AWS Reference Architecture Pelias Web client Mobile client Mapbox 2 3 4 5 6 7 8 9 10 Apache Airflow is hosted on an Amazon Elastic Compute Cloud (Amazon EC2) instance to orchestrate the entire workflow for this application. Point-of-Interest (POI) data from Safegraph is combined with data from OpenStreetMap (OSM) into Amazon Relational Database Service (Amazon RDS) PostgreSQL database. Polygon data from both Safegraph and OSM is indexed by H3 hexes via an Amazon EMR process. Telemetry data from Veraset mode and Xis ingested daily and get indexed on H3 hexes. The processed POI and telemetry data get saved in a centralized Amazon S3 bucket. All the processed data is combined using H3 indexes and geospatial functions provided by Amazon Athena. An EMR process ingests aggregated data from the Athena flow. The EMR process creates GeoJSON data and saves it back to the S3 bucket. The GeoJSON data is loaded in the map layer of Mapbox and Amazon DynamoDB. The front-end app is a react application deployed using AWS Amplify. The client calls Mapbox for map layers. Amazon API Gateway has two AWS Lambda functions configured
