Swami Sivasubramanian is the vice president of machine learning at AWS. He joined Amazon in 2005 as an engineer and since then built over 40 AWS services along with his team. His expertise in cloud computing and machine learning has helped him rise through the ranks through these years. At the ongoing AWS re:Invent 2021 event, Sivasubramanian presented some of the most intriguing launches from AWS on the machine learning front. We list some of the most significant ones.

Amazon DevOps Guru for RDS

It is a new machine learning-based capability for Amazon relational database service (RDS). It can automatically detect and diagnose database performance and operational issues, helping users resolve bottlenecks in much less time. This new feature builds on the existing capabilities of DevOps Guru to provide remediation recommendations to a variety of database-related problems. The DevOps Guru for RDS notifies developers and DevOps engineers, provides information and details on the problem, and suggests intelligent remediation recommendations.

AWS Database Migration Service (DMS) Fleet Advisor

Sivasubramanian introduced AWS Database Migration Service (DMS) Fleet Advisor to help make it easier and faster to get data to the cloud and match with the corresponding database service. It automatically builds an inventory of on-prem databases and analytics services by streaming it to Amazon S3. AWS’ team then analyses the data to match it with the appropriate amount of AWS Datastore. This helps in providing customised migration plans.

This approach is cheaper because users no longer have to rely on third-party consultants to move the data. It also makes it a lot easier to modernise the data infrastructure with powerful relational databases.

New SageMaker Features

AWS CEO Adam Selipsky introduced SageMaker Canvas, a no-code platform, to build machine learning models and generate accurate predictions in his keynote. AWS was not done with SageMaker just yet; in his keynote presentation, Sivasubramanian announced a few more features to SageMaker:

SageMaker Ground Truth Plus: It is a service that uses an expert workforce to deliver high-quality training datasets faster. It uses a labelling workflow, including ML techniques for active learning, machine validation, and pre-training.

SageMaker Inference Recommender: It is a tool to help users choose the best available compute instance to deploy ML models for better performance and lower costs. It automatically selects the appropriate compute instance type, count, model optimisations, and container parameters.

SageMaker Serverless Interface: It allows easy deployment of ML models for inference without the need for configuration and management of the underlying infrastructure.

SageMaker Training Compiler: It uses GPU instances more efficiently, which helps in accelerating the training of deep learning models by up to 50 per cent. It concerns deep learning models from high-level language representations to hardware optimised instructions.

SageMaker Studio Lab: It is a free service for developers to learn machine learning tools and techniques. With this new platform, customers will be able to focus on the data science aspect of machine learning without setting up or configuring any infrastructure. It is based on JupyterLab web application, and users are allowed to leverage any framework like PyTorch, MxNet, TensorFlow, NumPy, etc. Another advantage of using SageMaker Studio Lab is the integration with Github that enables customers to open, view, and edit any notebook.

Amazon Kendra Experience Builder

Amazon Kendra is a machine learning-powered intelligent search service. To this, in his keynote speech, Sivasubramanian announced three new features:

Amazon Kendra Experience Builder: It is a low/no-code platform for customers to deploy a fully functional and customisable search experience without having any prior machine learning experience. It delivers an intuitive visual workflow to build and launch a Kendra-powered search application on the cloud. It comes with AWS Single Sign-On (SSO) integration that supports popular identity providers like Azure AD and Okta while delivering secure end-user SSO.

Amazon Kendra Search Analytics Dashboard: It helps administrators and content creators understand how end users find relevant search results, their quality, and the gaps in the content. It is to help better understand the quality and usability metrics across one’s Kendra-powered search applications. This dashboard provides a snapshot of how users are interacting with search applications which could be viewed in a visual dashboard in the console.

Amazon Kendra Custom Document Enrichment: It allows users to build a custom ingestion pipeline to pre-process documents before they are indexed into Kendra. The enrichment is performed by simple rules that can be configured into the console or by invoking functions from Amazon Lambda.