The OpenSwarm project develops an automated system that monitors and counts motorboats in a protected marine area (PMAs), and tracks their speed to help manage underwater noise pollution. We therefore developed Amilcar, a solution to collect audio samples to be used as training data for an embedded Neural Network solution for boat detection and localization. Our objective is to record three synchronized audio files, each capturing the passage of a motorboat from a different location. We have three identical recording setups, each composed of Raspberry Pi, a hydrophone, GPS, and an analog-to-digital converter. For each audio file, a timestamp and GPS location are injected every second. We also have a setup on the motorboat, composed of Raspberry Pi and GPS, to log its position. We deployed Amilcar in the Etang de Thau, in the South of France. Each recording setup was placed in a waterproof container attached to a buoy, and the three buoys were positioned in a triangle formation, with the motorboat passing through it. In the end, we successfully obtained three audio recordings.