Skip to main content

Cloud Platforms

I build systems on AWS, Azure, and GCP, so these notes aren't introductions to any one provider. They're a record of what I've learned actually building and operating things on them, once a model or prototype has to handle real users, data, identity, networking, deployments, and cost.

This section holds three kinds of notes:

Platform-specific: things genuinely specific to one provider (services, APIs, identity model, deployment mechanics) and the odd provider-specific problem.

Cross-cloud: things that look similar across providers but aren't, like identity boundaries, private networking, observability, cost, and how workloads move (or don't) between clouds.

Architecture and decisions: the bigger calls, like managed services vs. containers, where to run inference, and how much portability is worth paying for.

Platforms

  • AWS: the oldest and largest provider, with the widest catalog of services.
  • Azure: Microsoft's cloud platform, tied closely to enterprise tooling and the wider Microsoft ecosystem.
  • GCP: originated Kubernetes, strong in data analytics and ML infrastructure.