AWS re:Invent Machine Learning Round-Up

AWS re:Invent Machine Learning Round-Up

By Kim Blomgren
December 14, 2020

Machine Learning (ML) is growing quickly.  In the first ever AWS ML keynote, Dr. Swami Sivasubramanian, VP of Amazon Machine Learning, kicked off by announcing that more than 100,000 customers now use AWS for ML use cases, and that AWS launched over 250 new ML features this year. Illustrating how the innovation will continue in 2021, AWS unveiled several new ML features and services at this year’s re:Invent.

To get started with AWS Machine Learning, download our paper

Amazon SageMaker

Amazon SageMaker was a central component of both AWS CEO Andy Jassy and Dr. Sivasubramanian’s keynote. AWS is committed to supporting every major ML framework with it and Jassy underscored that AWS will continue to support each of the major frameworks to give developers and data scientists the flexibility to build what they need. Specifically, AWS made several additions to the Amazon SageMaker family, including:

  • Amazon SageMaker Clarify is an exciting step forward in mitigating bias in ML. Designed to detect bias across the ML workflow — from training data to data labeling and selection to model drift — Clarify seeks to help ML developers and scientists address bias and provide greater transparency. Mitigating bias helps scientists create better models and encourages end-users to make better decisions based on those models. Clarify integrates with the SageMaker toolset.
  • Deep Profiling for SageMaker Debugger is a new solution that helps developers identify bottlenecks and maximize resource utilization for training. With it, users can visualize different system resources and adjust them based on recommendations to save developer resources.
  • Amazon SageMaker Edge Manager works to manage and monitor ML models across fleets of edge devices. Applies specific optimizations to run models faster while giving you visibility into the performance of your models on each device in your fleet.
  • Amazon SageMaker Pipelines for easy-to-build CI/CD pipelines. A CI/CD service for ML, SageMaker Pipelines define each step of the ML workflow and pipelines manages the dependencies between the steps and automates it for you. Offer base pipelines that you can use or tailor to your needs.
  • Amazon SageMaker Data Wrangler aggregates and prepares ML features by importing your data, recognizing it, recommending the best transformation to make — combining features for composite features when needed – previewing the transformations and then applying them to the whole data set to make ML data preparation much easier. Data Wrangler converts your data into something the model understands, prototypes it to make sure it works and applies it to the entire data set, greatly reducing resources needed for ML data preparation.
  • Amazon SageMaker Feature Store addresses issues around model sharing and storage. Feature Store seeks to make it easier to share, find and organize your ML features for training and inference with a new ML feature repository.

Machine Learning Integrations

AWS also unveiled a new emphasis on making ML more accessible to data analysts with several announcements that closely integrate SageMaker with AWS data stores and BI tools:

  • Amazon Aurora ML allows you to add ML predictions to applications directly with SQL.
  • Amazon Athena ML applies ML based predictions to S3 queries, allowing you to mine deeper insights from S3 data.
  • Amazon Redshift ML integrates SageMaker Autopilot and Redshift which makes it easier for data scientists to apply ML on that data. Once data is trained, the model can be applied directly from the dashboard.
  • Amazon Neptune ML runs predictions for graph applications by selecting the graph data needed for training and providing templates for developers to alert developers to customize ML models for certain scenarios.
  • In addition to integrations with AWS solutions, Dr. Sivasubramanian unveiled that AWS has integrated SageMaker with Domo, Sisense, and Qlik, with Tableau and Snowflake integrations coming in 2021.

Now in preview is Amazon QuickSight Q, a new AWS ML natural language question and answer service. Users can ask Amazon QuickSight Q any question using natural phrases and receive an answer in seconds. Using sophisticated natural language parsing, Q will provide the answer using all your data that you connect to it.

Addressing Machine Learning Problems

Many customers ask Amazon if they can simply solve their ML problems for them. AWS does so through a wide variety of AI services – from vision to speech and industrial AI. For example:

  • Amazon Lookout for Metrics uses ML to detect anomalies for metrics with root cause analysis, detecting unexpected changes in metrics with high accuracy by applying the right algorithm to the right data. It launches with 25 built-in connectors to services like Salesforce and Zendesk.
  • Amazon DevOps Guru uses ML to identify issues before they impact customers. It anticipates operational issues like misconfigured alarms, resources approaching their limits, overutilization, memory leaks, and more to catch them before they cause major issues.
  • Amazon HealthLake helps healthcare organizations store, transform, and analyze health and life sciences data at scale. Currently in preview, the new ML solution helps the healthcare industry translate medical data to identify trends, identify anomalies and infer other medical intelligence.

Andy Jassy covered many additional industry-specific services that utilize ML, like Amazon Monitron and Amazon Lookout for Equipment for manufacturers as well as a host of new functionality for Amazon Connect.

As AWS re:Invent continues throughout the month, stay tuned as we share additional news announcements from the show.

Don’t want to miss it? Subscribe to our blog today.

Written by Flux7 Labs

Flux7, an NTT DATA Company, is the only Sherpa on the DevOps journey that assesses, designs, and teaches while implementing a holistic solution for its enterprise customers, thus giving its clients the skills needed to manage and expand on the technology moving forward. Not a reseller or an MSP, Flux7 recommendations are 100% focused on customer requirements and creating the most efficient infrastructure possible that automates operations, streamlines and enhances development, and supports specific business goals.