AWS re:Invent 2020 Recap: Machine Learning Keynote
While this year’s AWS re:Invent may be entirely virtual, AWS has not disappointed this Data Scientist in the slightest. The long list of new releases in AWS’ machine learning stack will undoubtedly benefit all users — ranging from novices to experts in the field.
During Tuesday’s Machine Learning Keynote, Swami Sivasubramanian, VP of Amazon Machine Learning, structured his message around three tenets — which, together, give builders the freedom to invent. Under each tenet, he announced the new releases for machine learning, while also explaining how they fit together with the event’s other announcements — helping the audience weave together the bigger picture.
Provide firm foundations
The first tenet, “Provide firm foundations,” was the basis for the first announcement: faster distributed training with Amazon SageMaker. Using Habana Gaudi processors from Intel, AWS will soon offer EC2 instances built for machine learning (ML) training — yielding a performance increase of up to 40% over current GPU-based EC2 ML training instances for training deep learning workloads. This helps provide a firm foundation upon which developers can build and deploy machine learning, faster, while also reducing costs.
Create the shortest path to success
The second tenet, “Create the shortest path to success,” got the audience (particularly football fans like me) excited, as Sivasubramanian shared how AWS and the NFL are collaborating to achieve game and player simulation that can predict, treat and ultimately prevent player injury.
In the spirit of creating the shortest path to success for critical projects like this, AWS announced Amazon SageMaker Data Wrangler. This is a tool that I’m particularly eager to experiment with, as AWS suggests it’s a huge time saver for data transformation and discovery. I’m also pleased to see that Data Wrangler will soon integrate with Snowflake, MongoDB and Databricks — historically, AWS required AWS databases to seamlessly leverage their tools.
Another time saver is Amazon SageMaker Clarify, Amazon’s bias detection tool across the entire ML workflow. Not only does bias detection save time, if done well it improves overall model quality, flagging any drift that may occur as models age.
The next SageMaker release that I plan to utilize as an education tool is model profiling for Amazon SageMaker Debugger. This capability maximizes resources for training, GPU, CPU, network and I/O memory, by analyzing resource utilization and then making recommendations on how to adjust. (Wow!) We also learned about Amazon SageMaker Edge Manager, a new feature that manages and monitors machine learning models across fleets of smart devices up to 25 times faster when compared to hand-tuning said models.
Expand machine learning to more builders
The third tenet, “Expand machine learning to more builders,” is a principle I’m particularly passionate about. AWS has attempted to achieve this by releasing Amazon Redshift ML — which I’m enthusiastic to test out. It’s crucial to be able to more-easily experiment and deploy machine learning models — however, I’m wary of the suggestion that certain experts are no longer required in the process. Selecting, refining and deeply understanding a model is fundamental to extracting the most value possible from the output.
As a launch partner for Redshift ML, Rackspace Technology can help you make the most of this new feature:
“At Rackspace Technology we help companies elevate their AI/ML operations. We’re excited about the new Amazon Redshift ML feature because it will make it easier for our mutual Redshift customers to use ML on their Redshift with a familiar SQL interface. The seamless integration with Amazon SageMaker will empower data analysts to use data in new ways, and provide even more insight back to the wider organization.”
General Manager for Data Solutions, Rackspace Technology
AWS continues to pave the way for streamlined development and deployment. I’m thrilled to see the increasing number of capabilities across platforms. As a Data Science consultant, I am constantly interacting with different frameworks and infrastructure. AWS is my go-to solution for model development and as systems are increasingly compatible, the more I can focus on model refinement.
This year’s re:Invent is a three-week event, all virtual, and free. To watch the event live or view recordings, register here and don’t forget to visit the Rackspace Technology virtual booth to learn more about our new AWS solutions and enjoy an immersive interactive experience.
Google Cloud Next ’23 Highlights— AI and Beyond
September 14th, 2023
Why You Need an MLOps Framework for Standardizing AI and Machine Learning Operations
September 12th, 2023
Mastering Multicloud: The Power of Cloud-Neutral Data Storage
August 31st, 2023