Over the last ten years, virtual assistants like Alexa, Google Assistant, and Siri have become mainstream; hundreds of millions of users are setting their timers, controlling music, checking the weather, etc., on a daily basis. In this talk, we will provide a high-level overview of speech recognition technology (cloud-based vs. on-device, far-field, dictation vs. domain-specific, always-on listening, etc.) and then focus on on-device voice recognition approaches, with some code examples, as well as various tips on tuning the performance.