We are an human powered audio/video transcription service and have lots of training data that can be used for training a speech recognition system. An ASR-as-a-Service in the cloud kind of platform where we can use our data to continuously train and improve the models would be very useful for us.