15
Enabling machine
learning inferencing
at scale
Amazon Robotics develops sophisticated machinery and
software to optimize efficiency in Amazon fulfillment centers.
Opportunity
As a purveyor of cutting-edge technologies, Amazon Robotics
has long known that using artificial intelligence and machine
learning to automate key aspects of the fulfillment process
represented extraordinary potential gains—so in 2017, it
devoted teams to accomplishing just that.
Solution
As the company iterated on its machine learning project,
it turned to AWS and used Amazon SageMaker to create
solutions. Looking to reduce time-consuming bin scanning,
Amazon Robotics built the Intent Detection System, a deep-
learning-based computer vision system trained on millions of
video examples of stowing actions.
Results
• Easier management: Amazon SageMaker freed the Amazon
Robotics team from the difficult task of standing up and
managing a fleet of GPUs for running inferences at scale
across multiple regions
• Cost savings: The solution saved the company nearly 50
percent on machine learning inferencing costs
• Greater productivity: Using the machine learning-
powered solution, Amazon Robotics unlocked a 20 percent
improvement in productivity
• Scalability: Amazon SageMaker enabled the system
to scale horizontally during its rollout, and the team is
confident it can handle peak inference demands
"Amazon SageMaker doesn't just manage the hosts we use
for inferencing, it also automatically adds or removes hosts
as needed to support the workload."
Eli Gallaudet, senior software manager, Amazon Robotics
SUCCESS STORY: ACCELER ATING INNOVATION