Increasingly, we can use sound to tell us about the presence of threatened animals, but also learn about the health of ecosystems and animal populations, and learn what kinds of interventions are most effective. In this talk, we'll see how bird song models and flexible machine learning tools developed by the Google Perch team are being used to understand coral reef health, monitor bird abundance, discover new populations, and measure reproductive success in threatened species. By reducing the development time for new classifiers to less than a day, we allow experts to answser a much broader range of questions than ever before