NucleusResearch.com
3
Document Number: T176 December 2019
K E Y B E N E F I T S
With built-in algorithms, GPU-accelerated inferences, and fully managed training
environments, one of the primary benefits delivered by Amazon SageMaker is a shortened
development cycle, leading to cost savings and faster machine learning adoption. As a fully
managed service, the complexity is significantly less than a fully homegrown machine
learning ecosystem, so development teams can be optimized for agility; in some cases
developers were redeployed to other value-add tasks or business areas as a result of the
decreased complexity and reduced overall workload associated with machine learning tasks.
Since 2018, Nucleus has interviewed over 50 organizations encompassing over 400 unique
machine learning projects on AWS. These interviews have allowed us to learn about the
customers' machine learning use cases, the benefits of partnering with AWS, and the value
delivered by tools like Amazon SageMaker.
FASTER MODEL DEVELOPMENT AND DEPLOYMENT
With Amazon SageMaker, users can select commonly used algorithms that are pre-built,
allowing them to build models faster and begin training and making inferences much more
quickly. Additionally, dependencies between modules and files can be managed
automatically, eliminating tedious administrative work and reducing the opportunity for
human error. Customers interviewed by Nucleus reported reduced time to inference (the
time from model creation until it is trained and tuned to produce predictions on live data) by
33 to 50 percent. Multiple models can be in-progress simultaneously, allowing organizations
to scale up their existing machine learning initiatives and magnify the capabilities of the
internal data science and development teams.
COST SAVINGS
Customers realized cost savings from outsourcing the management of their machine
learning infrastructure to AWS. No longer having to purchase, configure, and manage on-
site infrastructure for enterprise-scale machine learning produces significant savings from
avoided hardware and IT staff expenses, with some companies able to reduce their spend
Developer productivity increased by 20 to 25
percent by automating administrative tasks
and accelerated model training