About

I build and run AWS platforms centered on EC2 capacity and containerized workloads, delivered through Infrastructure as Code. My day-to-day work is architecture and operations: making environments repeatable, scaling predictable, and costs explainable. I prefer systems that are automated, observable, and easy to operate under change.

I shipped my first e-commerce project in 1999, and that long runway still shapes how I think: systems should behave under stress, not only in demos.

Up to around 2016, a big part of my work was designing and tuning web and MySQL clusters. Much of that happened on providers like Hetzner, OVH, and Linode, where scaling was achievable but typically required planning and real lead time.

Moving into AWS changed the constraints. Scaling vertically or horizontally became close to immediate. The trade-off is money for time, but when the requirement is "yesterday", or when you need to absorb a short-lived spike, that speed matters.

Today I use both Terraform and CloudFormation depending on the task, and I typically prefer Terraform for larger or more complex projects because it scales well as an engineering system.

On the container side, I work with ECS and EKS when they fit, and I'm comfortable operating both-but I keep the focus on the fundamentals: container lifecycle, safe rollout patterns, workload isolation, and aligning scaling behavior with the underlying EC2 reality.

The AWS support layer is part of daily work-IAM, networking, CloudTrail, Athena, and CloudWatch-but I treat it as engineering, not background noise. In practice, CloudWatch is often the biggest recurring expense, so observability is something I design deliberately: collecting the signals that matter, controlling retention and volume, and keeping telemetry useful without becoming a cost leak.

More and more of my recent work supports private AI platforms and pipelines-from data preparation through training/fine-tuning and evaluation to production rollout and ongoing tuning. I also use AWS Bedrock occasionally and I'm actively evaluating where it fits best in real architectures.

For work, collaboration, or questions: contact@kalexandr.com