Once most advanced supervised or unsupervised machine learning models have been optimized and deployed, the next step for more powerful diagnostic and predictive analytics may require deep learning models and neural networks. Most trained neural network models for image and speech recognition are already available by major cloud service providers (e.g. AWS Lex) as an Artificial Intelligence service for developers. However, any group developing new solutions for their application can encounter situations where these AI solutions are not applicable. We have seen such a use case (e.g IOT energy sensors) arise where no additional feature engineering was possible and the next solution would be use a feed forward neural network for better prediction accuracy. We can definitely support any development effort to optimize a neural network layer architecture, and train it under offline evaluation. We can also work to develop the final production model that needs to be periodically updated with an new training data on a distributed cluster. The need for robust AI models is not limited to any one sector and we have seen the need for model to perform document and text classification (e.g. Pharma Marketing BI tools) come up as well that require a deeper neural network architecture (e.g. recurrent neural networks) to be developed and tested.