Bratin Saha, VP & GM, Machine Learning Services, Amazon, connects virtually with Dave Vellante for AWS re:Invent 2020 #theCUBE #reInvent #AWS https://siliconangle.com/2020/12/10/amazon-sagemaker-grows-ever-more-powerful-as-machine-learning-models-head-for-the-edge-reinvent/ Amazon SageMaker helps lead machine learning models to the edge VIDEO EXCLUSIVE BY BETSY AMY-VOGT Machine learning has experienced an incredible increase in usage in the past couple of years. In 2017, deploying machine learning models was considered extremely difficult, something only major organizations had the resources to consider. Then, Amazon Web Services Inc. released the Amazon SageMaker machine learning service. “Customers saw it was much easier to do machine learning once they were using tools like SageMaker,” said Bratin Saha (pictured), vice president and general manager of machine learning services and Amazon AI at Amazon. “Machine learning was no longer niche; machine learning was no longer a fictional thing. It was something that was giving real business value.” Saha spoke with Dave Vellante, host of theCUBE, SiliconANGLE Media’s livestreaming studio, during AWS re:Invent. They discussed trends in machine intelligence learning and artificial intelligence and how they’re supported by Amazon SageMaker. (* Disclosure below.) SageMaker makes machine learning accessible to every business As customers saw the value in machine learning, they went from deploying tens of models to deploying hundreds of thousands of models, making SageMaker one of the fastest-growing services in AWS history, according to Saha. “Today, we have one customer who is deploying more than a million models,” he said. But as usage grew, so did the problems customers were reporting. So, AWS grew SageMaker in response. More than 50 new capabilities have been added to SageMaker in just the past year, and re:Invent 2020 saw the addition of nine more. Running through these in his discussion with theCUBE, Saha described each in detail and gave key use cases. For example, SageMaker Data Wrangler is an addition to SageMaker Studio, a fully integrated environment for machine learning. By giving customers the ability to do data preparation in the same service as machine learning, it reduces the time spent cleaning data. “With a few clicks you can connect to a variety of data stores … and do all of your data preparation,” Saha said. “You get your data in; you do some interactive processing. Once you’re happy with the results of your data, you can just send it off as an automated data pipeline job. It’s the easiest and fastest way to do ML and take out that 80% [of time wrangling data].” Another new service is Amazon SageMaker Clarify, which provides visibility and transparency into the modeling process in order to eliminate unintentional biases. “[Clarify] helps to convert insights that you get from model predictions into actionable insights because you now know why the model is predicting what it is predicting,“ Saha said. Distributed Training on Amazon SageMaker increases the efficiency for customers training really large models, some of which now have billions of parameters, according to Saha. The new capability uses two techniques, model parallelism and data parallelism. The first enables these massive models that could take weeks to train on a single graphics processing unit to be trained in parallel across multiple GPUs. The second takes models that are too large to fit in the memory of a single GPU and automatically distributes them across multiple GPUs. These capabilities are “making it much faster and much easier for customers to work with large models,” Saha said. Bringing machine learning to edge devices and helping fulfill Amazon’s mission of “AWS Everywhere” is Amazon SageMaker Edge Manager. Optimizing models so they can run on an edge device increases the benefit by 25x, according to Saha. Not only does Edge Manager enable local inference, but it helps maintain model quality. “Once deployed it monitors quality of the models by letting you upload data samples to SageMaker so you can see if there is drift in your models or any other degradation,” Saha said. For complete descriptions and discussion on all the new Amazon SageMaker capabilities, here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of AWS re:Invent. (* Disclosure: Amazon Web Services sponsored this segment of theCUBE. Neither AWS nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)