The AWS re:Invent conference announced numerous tools and services for developers in 2019. This year, the developers at AWS paid special attention to machine learning development. In this article, we list down the top announcements on machine learning services at AWS re:Invent 2019.
(This list is in alphabetical order)
1| Augmented AI
Amazon Augmented AI or A2I provides built-in human review workflows for common machine learning use cases, such as content moderation and text extraction from documents, which allows predictions from Amazon Rekognition and Amazon Textract to be reviewed easily. This feature makes it easy for building and managing human reviews for machine learning applications.
The benefits of using Augmented AI are:
- Easily Implement Human Review of ML Predictions.
- Options to Work With the Choice of Human Reviewers
- Easily Integrate With Any Application.
Amazon CodeGuru is an ML service for automated code reviews and application performance recommendations. This service helps as well as recommends a developer in order to find and fix code issues such as resource leaks, potential concurrency race conditions, and wasted CPU cycles. Currently, CodeGuru supports Java applications and support for other programming languages will soon be released.
3| Contact Lens
Contact Lens for Amazon Connect is a set of ML capabilities which are integrated into Amazon Connect. It uses NLP and speech-to-text techniques to transcribe contact centre calls to create a fully searchable archive and surface valuable customer insights. Contact Lens helps the contact centre supervisors to better understand the sentiment, trends, and compliance risks of customer conversations to effectively train agents, replicate successful interactions, and identify the crucial company and product feedback.
4| Fraud Detector
Amazon Fraud detector is a fully-managed service which helps in identifying fraudulent online activities such as creating fake accounts, fraudulent online payment, among others. The detector overcomes these challenges by using user data and machine learning techniques. Some of the benefits of Amazon Fraud Detector are mentioned below
- Build high-quality fraud detection ML models faster: It provides templates that a user can use to easily create ML models to identify potential fraud without writing any code.
- Predicts Risk in the Information: The Fraud Detector predicts risk in the information while creating an account to identify fraudsters.
- Built-in Online Fraud Expertise: The pre-built machine learning model templates in Amazon Fraud Detector helps in detecting the fraudsters and stopping them to fraud AWS and Amazon.com.
Amazon DeepComposer is specifically designed to educate the developers which include tutorials, sample code, and training data to get started building generative AI models. DeepComposer includes Generative AI techniques which pit two different neural networks against each other in order to produce new and original digital works based on sample inputs. With AWS DeepComposer, a user can easily train and optimize GAN models to create original music.
Amazon Kendra is an enterprise search service which is powered by machine learning. The service delivers intuitive natural language search capabilities to the websites and applications so that the end-users are able to find their specific information. Some of the benefits of this service are mentioned below:
- Get Instant Answers: This service uses natural language questions which makes it easy to get the answers instantly.
- Data into Centralised Location: Kendra lets a user to easily add content from file systems, SharePoint, intranet sites, file-sharing services, and more, into a centralised location.
- Constantly Improving Search Results: The machine learning algorithm in Kendra’s keeps learning and is getting better over time search.
7| SageMaker Operators for Kubernetes
SageMaker Operators for Kubernetes are operators which can be used to train machine learning models, optimise hyperparameters for a given model, run batch transform jobs over existing models, and set up inference endpoints. This service helps the developers to use Kubernetes to train, tune, and deploy machine learning (ML) models in Amazon SageMaker.
The SageMaker Operators on the Kubernetes cluster can be installed in Amazon Elastic Kubernetes Service (EKS) to create SageMaker jobs natively using the Kubernetes API and command-line Kubernetes tools such as ‘kubectl’. In one of our articles, we already discussed how AWS Sagwmaker and Kubernetes integration will help ML developers.
8| SageMaker Studio
The Amazon SageMaker Studio is the first visual integrated development environment (IDE) for machine learning in AWS. It provides a single, web-based visual interface where a user can perform all ML development steps required to build, train, and deploy models. SageMaker Studio also enables one-click sharing of notebooks including the popular Jupyter Notebook
The Amazon SageMaker Debugger is one of the important services announced this year at re:Invent. It makes the training process more transparent by automatically capturing real-time metrics during training such as training and validation, confusion matrices, and learning gradients to help improve model accuracy.
10| SageMaker AutoPilot
Amazon SageMaker Autopilot is the industry’s first automated ML capability which gives a user complete control and visibility into the ML models. With a few clicks, the Autopilot automatically inspects raw data, applies feature processors, picks the best set of algorithms, trains, and tunes multiple models, tracks their performance, and then ranks the models based on performance.
This time, the developers at Amazon brought a lot of new services to their SageMaker. Apart from these 10 services announcements at AWS re:Invent, there are few more services on machine learning announced:
- SageMaker Model Monitor which allows developers to detect and remediate concept drift
- SageMaker Notebooks provide one-click Jupyter notebooks along with one-click sharing
- SageMaker Experiments helps a user to manage iterations by automatically capturing the input parameters, configurations, and results, and storing them as ‘experiments’