So, who am I?

I'm a senior engineer, 11+ years into platform engineering and distributed systems, mostly inside regulated financial and public-sector environments. I design and run the common infrastructure other teams build on - multi-tenant cloud foundations, GitOps self-service, secure data layers, the kind of systems that have to hold up under regulatory scrutiny and not just a demo.

Most of what I write about these days lives at the intersection of that and AI-native development. kubemend is an LLM agent harness for Kubernetes remediation, hand-written with no agent frameworks, that diagnoses incidents from live cluster state and can only act by opening draft pull requests - a verification pipeline the model doesn't control re-checks every change before it's ever called "done." I've written up what building it actually taught me - costs, bugs, an adversarial prompt-injection test - in a short series of articles.

Outside of platform work I like building things end-to-end on my own. Rnkd is a consumer product I shipped solo - design, mobile frontend, backend, infrastructure, distribution - and rgw-model is a multi-agent reinforcement-learning environment, closer to my background in an MSc on multi-agent RL. Both, and a few other things, are on the projects page.

I'm a Kubestronaut (all five CNCF Kubernetes certifications) and an AWS Certified Solutions Architect - Professional.

You can find more of my code on GitHub, or say hello at maciej@stepkowski.dev, on LinkedIn, or on X.