As we have seen in this book, there are various SDKs available that will help programmers create mobile ML programs:

• TensorFlow Lite
• Caffe2Go
• Core ML
• ML Kit
• Fritz

We have gone through the high-level architecture of these SDKs and also written sample mobile ML applications using each of the preceding SDKs in this book. 

There are opportunities for all in the following areas of mobile ML technology enablement:

• Just like hybrid mobile application development, there can be way for hybrid ML model development that could employ a common language to develop these ML models
• There are many issues in terms of model deployment and upgrading of the deployed live ML models in the field
• There are many things to improve in terms of monitoring the performance and usage of the ML models in the field
• There is a lot of work to be undertaken to support many ML algorithms in these SDKs
• Now, predominantly, only prediction and usage of models is being done from mobile; training can also be undertaken from mobile on-device training enablement