Amazon today launched SageMaker Data Wrangler, a new AWS service designed to speed up data prep for machine learning and AI applications. Alongside it, the company took the wraps off of SageMaker Feature Store, a purpose-built product for naming, organizing, finding, and sharing features, or the individual independent variables that act as inputs in a machine learning system. Beyond this, Amazon unveiled SageMaker Pipelines, which CEO Andy Jassy described as a CI/CD service for AI. And the company detailed DevOps Guru, QuickSight, and QuickSight Q, offerings that uses machine learning to identify operational issues, provide business intelligence, and find answers to questions in knowledge stores.
During a keynote at Amazon’s re:Invent conference, Jassy said that Data Wrangler has over 300 built-in conversion transformation types. The service recommends transformations based on data in a target dataset and applies these transformations to features, providing a preview of the transformations in real time. Data Wrangler also checks to ensure that the data is “valid and balanced.” As for SageMaker Feature Store, Jassy said that the service, which is accessible from SageMaker Studio, acts as a storage component for features and can access features in either batches or subsets. SageMaker Pipelines, meanwhile, allows users to define, share, and reuse each step of an end-to-end machine learning workflow with preconfigured customizable workflow templates while logging each step in SageMaker Experiments.
DevOps Guru is a different beast altogether. Amazon says that when it’s deployed in a cloud environment, it can identify missing or misconfigured alarms to warn of approaching resource limits and code and config changes that might cause outages. In addition, DevOps Guru spotlights things like under-provisioned compute capacity, database I/O overutilization, and memory leaks while recommending remediating actions.
Amazon QuickSight aims to provide scalable, embeddable business intelligence solutions tailored for the cloud. To that end, Amazon says it can scale to tens of thousands of users without any infrastructure management or capacity planning. QuickSight can be embedded into applications with dashboards and is available with pay-per-session pricing, automatically generating summaries of dashboards in plain language. A complementary AWS service called QuickSight Q answers questions in natural language, drawing on available resources and using natural language processing to understand domain-specific business language and generate responses that reflect industry jargon.
Amazon didn’t miss the opportunity this morning to roll out updates across Amazon Connect, its omnichannel cloud contact center offering. New as of today is Real-Time Contact Lens, which identifies issues in real time to impact customer actions during calls. Amazon Connect Voice ID, which also works in real time, performs authentication using machine learning-powered voice analysis “without disrupting natural conversation.” And Connect Tasks ostensibly makes follow-up tasks easier for agents by enabling managers to automate some tasks entirely.
Amazon also launched Amazon Monitron, an end-to-end equipment monitoring system to enable predictive maintenance with sensors, a gateway, an AWS cloud instance, and a mobile app. An adjacent service — Amazon Lookout for Equipment — sends sensor data to AWS to build a machine learning model, pulling data from machine operations systems such as OSIsoft to learn normal patterns and using real-time data to identify early warning signs that could lead to machine failures.
For industrial companies looking for a more holistic, computer vision-centric analytics solution, there’s the AWS Panorama Appliance, a new plug-in appliance from Amazon that connects to a network and identifies video streams from existing cameras. The Panorama Appliance ships with computer vision models for manufacturing, retail, construction, and other industries, supporting models built in SageMaker and integrating with AWS IoT services including SiteWise to send data for broader analysis.
Shipping alongside the Panorama Appliance is the AWS Panorama SDK, which enables hardware vendors to build new cameras that run computer vision at the edge. It works with chips designed for computer vision and deep learning from Nvidia and Ambarella, and Amazon says that Panorama-compatible cameras will work out of the box with AWS machine learning services. Customers can build and train models in SageMaker and deploy to cameras with a single click.
The slew of announcements come after Amazon debuted AWS Trainium, a chip custom-designed to deliver what the company describes as cost-effective machine learning model training in the cloud. Amazon claims that when Trainium becomes available in the second half of 2020, it’ll offer the most teraflops of any machine learning instance in the cloud, where a teraflop translates to a chip being able to process one trillion calculations a second.