Bring machine-learning to your projects with Amazon SageMaker. The cloud platform makes it easy to deploy machine-learning models in the cloud, on edge devices, and embedded systems. Use SageMaker to quickly create and train machine-learning models with multiple levels of abstraction.
SageMaker protects your data, encrypting the work while at rest and in transit. An SSL connection handles all requests. Encrypted S3 buckets and a KMS key protect models, data, notebooks, endpoints, and storage volume.
SageMaker does not charge for anything you don't use. No minimums or commitments. Billing is based on the number and types of instances used.
Transfer data and results in and out of SageMaker to use existing tools or take advantage of SageMakers end-to-end workflow.
SageMaker Autopilot makes creating machine-learning models less labour-intensive. Autopilot inspects data and determines the best algorithms to use on the model. It runs multiple scenarios, tracks results, and provides a performance ranking.
SageMaker supports PyTorch, TensorFlow, Keras, Deep Graph Library, Scikit-learn, Apache MXNet, and other frameworks. SageMaker Autopilot supports both R and Python. Algorithms supported at launch include Linear Learner and XGBoost. SageMaker supports distributed training, as well.
Share notebooks with one click using SageMaker. Included Jupyter notebooks let you adjust resources as needed and update in the background as you work. SageMaker comes with dozens of pre-built notebooks for a range of use cases.
Amazon SageMaker Experiments provide the tools need to organize and track variations of your models. Isolating and measuring the results from changing data sets, model parameters, and algorithm versions is a time-consuming process. SageMaker Experiment handles this automatically, capturing the varying parameters and results and storing them for review. From SageMaker Studio, you can explore and compare various iterations of your work.