By Hans Hsu, Project Solutions Architect, Chuan Kai International Co. LTD (CKmates)
The trend of ubiquitous artificial intelligence that began from 2017 to 2018, after two years of development has matured in 2019. How to choose a suitable solution for the business logic of the enterprise in the numerous AI services provided by the public cloud is undoubtedly a significant topic.
The AI services provided by the major public clouds that we can classify into three layers:
- Application Layer
- Platform Layer
- Infrastructure Layer
The application layer means that the enterprises directly use the out-of-the-box functions provided by the public cloud. Thus, Developers can develop without understanding complex mathematical theories and model concepts. They only use AI functions through API, merely focus on the workflow and business logic of the company instead of cooperating with data scientists and data engineers by a complicated training process.
The platform layer refers to the enterprise directly using the development platform provided by the public cloud, enabling enterprises to customize the unique AI features. For instance, AmazonSageMaker integrates the most common tools used by data scientists or data engineers into a service or SDK, can be united and can be scaled the computing performance horizontally; you can also control the model version and replace the appropriate model. In short, there are services available from data cleaning to training models, but the business logic and ownership of the model lies in the company. At the platform level, companies can focus on AI model development rather than managing infrastructure (e.g., operating system updates, hardware updates, etc.) and the installation of related packages, therefore, companies can develop AI features that better meet the business demands.
The infrastructure layer refers to the operation of the enterprise to customize the operating system or adjust the underlying parameters of the operating system. The public cloud also provides directly usable computing resources: Virtual machine (EC2), AWS provides flexible NVIDIA GPU (Elastic Graphics, Amazon Elastic Inference) or AWS's EC2 optimized for AI operations and pre-installed drivers and prevalent machine learning and deep learning framework, you can enjoy AWS's fast and efficient AI hardware resources directly after the operating system boot.
However, it's a considerably complex process that chooses the right solution among numerous services and integrates them into the enterprise. Ckmates, as an AWS MSP Provider, supports clients with customized solutions for the enterprise, selects appropriate AWS services according to the customer's current technology stack, integrates existing architecture, helps customers set milestones and grows with customers. For example, Assist in combining AI development and DevOps automation processes, AI computing resource configuration in the on-premise data center and the cloud, security of data used by cloud AI and assisting in the protection of public cloud technology. Under the framework provided by the public cloud, through the above-mentioned interactive use of several layers, whether it is a professional IT company or a non-IT professional company can easily select the appropriate AI development solutions from which to enhance the competitiveness of the enterprise.