AWS Proton is a fully managed delivery service that creates and manages standardized infrastructure and deployment tooling for developers and their serverless and container-based applications. It became generally available in June.

AWS Proton is designed for users to more easily provision, deploy and monitor microservices that are the foundation of modern container and serverless apps. It provides management tools, governance and visibility required for consistent standards and best practices for managing deployments, while also helping increase developer productivity, according to AWS.

AWS Proton has a two-pronged automation framework: Platform operators can use infrastructure as code to create versioned service templates that define and configure what’s needed to provision, deploy and monitor a service, and developers then can choose published stacks to rapidly build applications, knowing they are working with up-to-date, validated tools and infrastructure.

Two new features were added since AWS Proton’s launch in preview last fall. It now supports multi-account infrastructures, allowing platform operators to use it to help securely configure and manage their architectures across multiple AWS accounts. It also includes support for identity and access management condition context keys in AWS Proton APIs, allowing platform operators to designate which developers can create services based on template characteristics. Amazon Lookout for Vision, a machine learning service that identifies defects and anomalies in visual representations, was launched into general availability in February.

It uses AWS-trained computer vision models to analyze images and video streams to find flaws and anomalies in manufactured products or production processes, damage to vehicles or structures, and defects in silicon wafers or any physical item where quality is important, such as a missing capacitor on printed circuit boards, according to AWS.

Amazon Lookout for Vision uses a machine learning technique called “few-shot learning” to train a model for a customer using as few as 30 baseline images. It can process thousands of images an hour with no machine learning experience required on the part of the user.

The offering analyses the data and then reports images that differ from the baseline through the service dashboard or a “DetectAnomalies” real-time API. The service is advanced enough to maintain high accuracy with variances in camera angle, poses and lighting arising from changes in work environments, according to AWS, and it can use feedback to retrain the underlying model so it continuously improves. Amazon Elastic Container Service (Amazon ECS) is a cloud-based, fully managed container orchestration service that became available to all AWS customers in 2015. It now has a new extension that allows customers to deploy native Amazon ECS tasks in any environment: Amazon ECS Anywhere, which became generally available in May.

Amazon ECS Anywhere enables users to easily run and manage container-based applications on premises, including on virtual machines, bare metal servers and other customer-managed infrastructure. Customers get the same AWS-style APIs and cluster configuration management pieces on premises as in in the cloud. Amazon ECS Anywhere precludes them from having to run or maintain their own container orchestrators on-premises.

Benefits include reduced costs and complexity when running container workloads such as data processing at edge locations on a customer’s hardware with reduced latency, and in the cloud using one standardized container orchestrator, according to AWS.