Deploying MeMySelfI: Docker on AWS EC2 with a CI/CD pipeline I built myself
How my AI agent platform got from laptop to live production: containerised with Docker, running on a single EC2 instance, with a pipeline I assembled myself — and what operating it taught me.

MeMySelfI — my AI agent platform — only became a real product the day it stopped running on my laptop. The deployment I chose was deliberately unfashionable: one Docker image, one AWS EC2 instance, and a CI/CD pipeline I put together myself instead of reaching for a platform that hides the moving parts.
Containerising the services first forced the discipline everything else depends on: configuration through environment variables, no state hiding on the local filesystem, and a build that behaves the same on my machine as it does on the server. Once the image is trustworthy, EC2 is the simple part — pull, run, and keep the host boring.
The pipeline exists because manual deploys are where mistakes breed. Building it myself — rather than inheriting a template — meant I understood every step the day something failed, which it did, and I could fix the step instead of fearing the whole chain.
The lesson that stayed with me: production is a feature. Agent code gets the attention, but the Docker file, the pipeline and the restart behaviour are what users actually experience.