A valaha volt legnagyobb Nemzeti Menet

[0:00]Hello, my name is Joe and I'm a solutions architect here at AWS, focusing on aim and machine learning. And in this session, we're going to be talking about how to leverage Amazon SageMaker to easily develop, train, and deploy machine learning models. We're going to start off by talking about some of the challenges that customers face today when trying to do machine learning. We're then going to dive into how Amazon SageMaker can help solve those problems across the different phases of the machine learning workflow. And then we're going to wrap up with some customer success stories and some next steps. Now, as you're all aware, machine learning is eating the world, right? We're seeing it everywhere from self-driving cars to things like fraud detection, to things like personalization and recommendations. So businesses are adopting machine learning at a very rapid rate. However, there are some challenges with machine learning. And the first one is around the expertise needed to do machine learning, right? It's not just about knowing how to train a model, but you also need to know things like data engineering. You need to know how to deploy and operate your model. So there's a lot of different skill sets that are needed to be able to do machine learning at scale. The second challenge is around the cost and time involved in doing machine learning. And a lot of this has to do with things like infrastructure, right? When you're dealing with very large data sets, you need the compute to be able to process those data sets. You need the GPUs to be able to quickly train your models and iterate on those models. And then finally, you also need the ability to scale your models and integrate them into your business processes. And the third challenge is around the complexity of the machine learning workflow. And as you can see, there's a number of different steps involved, starting from data collection and preparation, to feature engineering, to model training, to model tuning, to model deployment and management. There's a lot of things involved, and historically, you had to piece together a number of different tools and services to be able to accomplish all of these different steps. So what is Amazon SageMaker? It's a fully managed service that helps you develop, train, and deploy machine learning models quickly. It provides all of the components that you need to be able to do machine learning at scale. And we're going to dive into each of those components as we go through the different phases of the machine learning workflow. So we're going to start off with the build phase and specifically talk about SageMaker Studio. SageMaker Studio is the first and only fully integrated development environment for machine learning. It provides a single pane of glass for you to be able to do all of your machine learning activities. So it's a web-based IDE that gives you access to things like Jupiter notebooks. It also gives you the ability to do things like experiment management, model debugging, and model monitoring. So it really is a single place for you to be able to do all of your machine learning activities. It also provides you with access to things like Amazon S3, which is our object storage service, which you can use to store your data sets. It also gives you access to things like Amazon Redshift, which is our data warehousing service, which you can use to query your data sets. And then finally, it gives you access to things like Amazon Athena, which is our interactive query service, which you can use to query your data sets stored in S3. So it really is a single place for you to be able to do all of your machine learning activities. Now, once you've prepared your data and you've done your feature engineering, the next step is to train your model. And SageMaker provides a number of different options for you to be able to train your models. The first one is around built-in algorithms. So SageMaker provides a number of different built-in algorithms that you can use to train your models. These are highly optimized algorithms that are pre-trained on large data sets. And you can use them to quickly train your models without having to write any code. The second option is around custom algorithms. So if you have your own custom algorithms, you can bring them to SageMaker and train them there. And then finally, the third option is around SageMaker jumpstart. SageMaker jumpstart provides a number of different pre-built solutions that you can use to quickly get started with machine learning. These are end-to-end solutions that include everything from data preparation to model training to model deployment. And you can use them to quickly get started with machine learning without having to write any code. Now, once you've trained your model, the next step is to tune your model. And SageMaker provides a number of different options for you to be able to tune your models. The first one is around hyperparameter tuning. So hyperparameter tuning is the process of finding the best hyperparameters for your model. And SageMaker provides a number of different strategies for you to be able to do hyperparameter tuning, including things like random search, grid search, and Bayesian optimization. The second option is around automatic model tuning. So automatic model tuning is the process of automatically finding the best model for your data set. And SageMaker provides a number of different algorithms for you to be able to do automatic model tuning, including things like random forest, gradient boosting, and neural networks. Now, once you've tuned your model, the next step is to deploy your model. And SageMaker provides a number of different options for you to be able to deploy your models. The first one is around real-time inference. So real-time inference is the process of deploying your model to an endpoint that you can use to get real-time predictions. And SageMaker provides a number of different options for you to be able to do real-time inference, including things like multi-model endpoints, which allow you to deploy multiple models to a single endpoint. The second option is around batch inference. So batch inference is the process of deploying your model to a batch transform job that you can use to get predictions on a batch of data. And SageMaker provides a number of different options for you to be able to do batch inference, including things like batch transform, which allows you to get predictions on a batch of data. And then finally, the third option is around SageMaker Edge. SageMaker Edge allows you to deploy your models to edge devices, such as IoT devices, and get predictions on those devices. So it really allows you to deploy your models to wherever your data resides. Now, once you've deployed your model, the next step is to manage your model. And SageMaker provides a number of different options for you to be able to manage your models. The first one is around model monitoring. So model monitoring is the process of monitoring your models for things like data drift and model drift. And SageMaker provides a number of different options for you to be able to do model monitoring, including things like model monitor, which allows you to monitor your models for data drift and model drift. The second option is around model explainability. So model explainability is the process of understanding why your model is making the predictions it is making. And SageMaker provides a number of different options for you to be able to do model explainability, including things like SageMaker clarify, which allows you to understand why your model is making the predictions it is making. And then finally, the third option is around model lineage. So model lineage is the process of tracking the lineage of your models, including things like the data sets that were used to train your models, the algorithms that were used to train your models, and the hyperparameters that were used to train your models. And SageMaker provides a number of different options for you to be able to do model lineage, including things like SageMaker experiments, which allows you to track the lineage of your models. So as you can see, SageMaker provides a number of different options for you to be able to manage your models. Now, let's talk about some customer success stories. And the first one is around NFL. So NFL is using SageMaker to power their next generation stats platform. And they're using SageMaker to do things like player tracking, game analysis, and fantasy football. So they're using SageMaker to really power their entire stats platform. The second customer success story is around 3M. So 3M is using SageMaker to accelerate their innovation. And they're using SageMaker to do things like material science, product design, and manufacturing. So they're using SageMaker to really accelerate their innovation across their entire organization. And then finally, the third customer success story is around T-Mobile. So T-Mobile is using SageMaker to improve their customer experience. And they're using SageMaker to do things like churn prediction, customer segmentation, and personalization. So they're using SageMaker to really improve their customer experience across their entire organization. Now, let's talk about some next steps. And the first one is around getting started with SageMaker. So if you're interested in getting started with SageMaker, you can go to our website and sign up for a free tier account. You can also access a number of different resources, including things like documentation, tutorials, and code samples. And then finally, you can also reach out to your account team or your solutions architect to get started with SageMaker. The second next step is around exploring the SageMaker features. So if you're interested in exploring the different SageMaker features, you can go to our website and learn more about the different features that are available. You can also access a number of different resources, including things like documentation, tutorials, and code samples. And then finally, you can also reach out to your account team or your solutions architect to learn more about the different SageMaker features. And the third next step is around attending our workshops and webinars. So if you're interested in attending our workshops and webinars, you can go to our website and sign up for our upcoming workshops and webinars. You can also access a number of different resources, including things like recordings, presentations, and code samples. And then finally, you can also reach out to your account team or your solutions architect to learn more about our workshops and webinars. So with that, I want to thank you for your time and I hope you enjoyed this session. And I look forward to seeing you at our next session.