Millions of accelerometers embedded in smart phones, laptops, and other consumer devices are capable of measuring large earthquakes as they spread across cities. Harnessing the data from these sensors to detect and estimate earthquakes is an Internet-scale problem involving real-time inference on a vast set of noisy sensor measurements. Few assumptions can be made about the data reported by each sensor: beyond differences in device hardware, each accelerometer is owned and operated by a different individual. To succeed, a seismic network must learn and adapt to these a priori unknown factors. The Caltech Community Seismic Network (CSN) employs online learning and decentralized anomaly detection to maximize network-wide seismic detection performance under constraints on the false positive rate and bandwidth usage. This demo showcases how the CSN uses phones (Android devices) and cloud computing (Google App Engine) to build a scalable, decentralized platform for data collection and inference. We will present two key technologies: First, we show the latest developments in CSN-Droid, an Android app that uses anomaly detection to report potential seimic measurements. Second, we show how the App Engine fusion center is used to collect sensor data, detect anomalous spatial events, visualize them on a map, and issue real-time alerts to registered Android devices about ongoing quakes. Participants will be able to interact with the end-to-end system by producing their own ``seismic activity'' using Android devices, and evaluate the alerts and visualizations produced by the fusion center.