In this talk, Marco, an engineer manager at Mozilla, explains how they have used machine learning to improve software engineering at Firefox. He starts by discussing the complexity of the Firefox browser and the large number of bug reports they receive. They introduced a new type field in Bugzilla to differentiate between defects and feature requests, using machine learning to classify existing bugs. They also used machine learning to automatically assign bugs to the appropriate team based on components. Additionally, they applied machine learning techniques to improve testing efficiency, by selecting relevant tests to run on patches using models trained on the relationships between patches and tests. They also trained models to predict the riskiness of a patch and detect spam in bug reports. Finally, Marco discusses a project on privacy-friendly translations in Firefox, where they used machine learning to perform translations locally on users' machines instead of using cloud-based services.