Amazon Web Services (AWS) is a highly evolved comprehensive cloud suite for clients who want a range of resources that can facilitate machine learning and deep learning projects. As a major cloud provider, AWS offers specialized machine learning and deep learning services to enable clients to build AI infrastructures.
According to its marketing materials, Amazon provides machine learning resources in three "layers of the AI stack" – the first one is framework tools, the second one API-driven services, and the third is machine learning platforms.
In terms of frameworks, Amazon offers "AWS deep learning AMIs" that include tools like MXNet and TensorFlow. Amazon describes an Amazon Machine Image or AMI as something that "provides the information required to launch an instance, which is a virtual server in the cloud." So the AMI is the vehicle for clients working with tools and building the systems that will support their enterprise work.
Amazon has also developed API-driven services that support machine learning projects initiated by clients. One is the "Lex" application that Amazon says is built off of the consumer product Alexa, and fulfills a similar role in a client machine learning process, dealing with natural language and allowing clients to build chatbots and other communications elements into their systems. Another API-driven service called "Polly" changes text into speech, and a third instance, "Rekognition," is an image processing utility.
Lastly, AWS offers machine learning platforms that clients can use to avoid overhead associated with in-house production of machine learning environments. Amazon Machine Learning and Apache Spark on Amazon Elastic MapReduce (EMR) are options for using AWS to initiate machine learning projects. The Amazon Machine Learning platform provides guidance that helps less tech-savvy users to develop machine learning skills. Amazon describes it as a "highly scalable" resource that can provide real-time productivity over a web-delivered platform. Clients can also use Apache Spark and apply it to Amazon EMR for Hadoop-related processing. This involves using a spectrum of Apache open-source tools to create a client’s desired functionality.
The ways in which AWS facilitates machine learning for clients are diverse. AWS tools work with Elastic Compute Cloud (EC2) services to offer this kind of "build your own" machine learning option for clients.