AI technology is expanding rapidly. The number of ultra-small constrained devices is too. Can these two trends be reconciled? This talk will examine keyword detection in earbuds, a real-world example of bringing AI technology to bear in a highly constrained platform. We will ask why doing so is beneficial and discuss some of the challenges encountered. We will take the audience through the stages involved in building a neural network-based detection system, including worked examples. We aim to raise awareness of the increasing need for collaborative neural networks that operate across the wider eco-system to unlock the full potential of end-point devices.