As companies take machine learning out of R&D and into production, they face a whole new set of challenges. Lukas Biewald explains why human in the loop, active learning, and transfer learning are all essential design patterns for making deep learning real.