Intro to AWS Machine Learning Services

AWS Machine Learning stack has 3 key layers

1. Framework & Infrastructure
2. Machine Learning Platforms
3. API Driven Services

For experts, it provides a baseline to create deep learning models, training them and doing inference and getting the data from the models into the production applications.

AWS also provides Optimized Instances and Machine Images (AMI — Amazon Machine Image) which are designed for experts to get started quickly using their preferred framework.

Deep Learning AMI has all the major frameworks like —

• Apache MXNet — This is great for recommendation engines
• Caffe & Caffe 2 — For Computer Vision Projects
• TensorFlow — Easy to build models

For those who are looking for fully managed platforms for building models using their data, AWS provides —

• Apache Spark on Amazon EMR (Elastic Map Reduce)
• SparkML

API based tools

• Amazon Recognition — Image Recognition
• Amazon Polly — Text to Speech
• Amazon Lex — Intelligent Conversational Interfaces (Same tech as Alexa)

Machine Learning is the power of an underlying architecture that builds on the quality of the underlying data.

So, to be good at machine learning we would need a Data Store that is scalable, available, secure and flexible. This data lake must handle extremely large datasets. For this AWS provides the following options

Amazon S3 (used as a Data Lake)

Data Analytics —

• Athena
• Redshift
• Redshift Spectrum

Amazon EMR (provides access to frameworks like)

• Apache Spark
• Presto
• Hive
• Pig

Most choose AWS because it provides all the capabilities needed to do machine learning which works together very well.