AWS

AWS, short for Amazon Web Services, is Amazon’s cloud computing platform. It rents out compute power, storage, databases, and hundreds of other infrastructure services over the internet, on a pay-as-you-go basis, so companies never have to buy or maintain their own physical servers. Launched in 2006, AWS now spans over 240 services across 39 geographic regions worldwide. For AI developers, AWS matters because it provides the compute and managed AI services, like Amazon Bedrock and SageMaker, that most production machine learning systems actually run on.

How AWS works

AWS delivers infrastructure as a service (IaaS), meaning you rent raw computing resources instead of buying hardware, and platform and software services built on top of that infrastructure.

• Compute: services like EC2 (Elastic Compute Cloud) give you virtual servers you can spin up or shut down in minutes, scaling to match demand.
• Storage: S3 (Simple Storage Service) stores any amount of data as objects, accessed over the internet and billed by usage.
• Serverless: AWS Lambda runs your code in response to events without you managing any server at all, billing only for the compute time actually used.
• Regions and availability zones: AWS infrastructure is split into geographic regions, each containing multiple isolated availability zones, so applications can stay online even if one data center fails.
• Identity and access management (IAM): every resource in AWS is protected by fine-grained permissions, controlling exactly who and what can access each service.

Everything in AWS is accessible through a web console, a command-line interface, or an API, which is what makes it possible to automate infrastructure entirely in code.

What you can build with AWS

• Generative AI applications: use Amazon Bedrock to call foundation models through a single API, adding retrieval-augmented generation with Knowledge Bases and safety controls with Guardrails, without hosting any model yourself.
• Custom ML pipelines: use Amazon SageMaker to train, fine-tune, and deploy your own machine learning models at scale, with managed infrastructure for experimentation and versioning.
• AI agents that take action: deploy agents through Bedrock AgentCore that can reason about a task, call external APIs, query databases, and complete multi-step workflows.
• Scalable web applications: run a full-stack app on EC2 or serverless Lambda functions, backed by managed databases, that automatically scale from a handful of users to millions.
• Document processing systems: extract structured data from unstructured documents like invoices, contracts, or medical records using Bedrock Data Automation.
• Global content delivery: serve static assets, videos, and APIs from edge locations close to your users, cutting latency for applications used worldwide.