Engineering and Scientist at SublimeAI keep developing innovative consumer applications in imaging, natural language processing, audio, sequence modelling, etc. With launch of each such application the SublimeAI engine continues to enhance and become more powerful.
SublimeAI Deep Learning Platform Architecture
Specializes in learning sequences. Uses Recurrent Neural Networks (RNN) and Long Short Term Memory (LSTM) networks
Specializes in learning spatial relationships. Uses Convolution Neural Networks behind the scenes.
On-line Learner / Model Updater
Deep learning models need to be continuously updated. This module automates the process of training a model and then replacing it in the inference engine so that the latest model is being used.
Deep learning models need to be ultimately deployed in production. This process is called inference. The inference engine at Sublime AI handles the process of serving a model either using an API as a web server or converting the model into a mobile friendly model file that can be then deployed to mobile application.
This component takes existing datasets and then creates training, test, and validation sets. It also prepares data for semi supervised learning by using sudo sampling.
This component is responsible for serving request over the web service. For instance it provides functions for things like train, test, sub-sample, upload, etc.
This component handles the mobile presentation layer. This has further components like
Sublime AI deployment comes in two flavors. One is on premise and the other is on cloud. The benefits of the on cloud model is that it is much more easier to develop, deploy and keep getting the benefits of the Sublime AI's engine. The benefit of the On Premise model is that its easy to integrate with data since there is no copy of data made, and sometimes because of security policies the on premise is a need not a want. Many companies now days use AWS, Azure, or Google Cloud as their data centers. Sublime AI can also be deployed as an app on these platforms. Which option we choose is a matter of discussion and time to market.