Hivemapper is a decentralized global map data network built by 10s of thousands of mapping devices. High-res sensors like RGB, Stereo Depth, GNSS, IMU, etc. feed sensor fusion and ML models at the edge. Data is automatically uploaded in near realtime over LTE or WiFi. Enterprise tech, mapping, auto, robotaxis, rideshare, and entertainment represent some of the customers consuming data today. APIs allow anyone to consume precisely extracted Map Features, HD map data, high-res street-level imagery, construction, and driving events for AV simulation. Tech-savvy customers develop and deploy software directly to our dashcams to get realtime data for things like change detection or visual semantic data mining. AI Fleet management tools drive value to large fleets of vehicles. Map AI Platform Every day, we process data from millions of KM from 10s of thousands of high resolution sensors deployed around the world. A symphony of different Sensor Fusion, ML, AI, 3D Reconstruction processes are performed at the edge and in our processing clusters, for which the Map AI Platform orchestrates. The Map AI Platform is the core data engine that powers our data customers, consumer products, and user-facing AI technology. These complex systems control and learn patterns about map data all over the world and enable us to to continuously evolve map data intelligence, semantic extraction, QA at scale. For example, we’ve recently been using the system to learn hyperlocalized road network patterns to automatically QA and correct noise from map conflation. We strongly believe in engineering design principles around abstraction and programmability that allow us to move quickly with minimal manual interventions. Our work is fast-paced, collaborative, and data-driven.