MK Michał Kajszczak infrastructure · reliability · AI agents

Taking on engagements from January 2027

Keep your infrastructure stable while AI agents start changing it.

I'm Michał, a platform and DevOps architect with more than ten years in production. I help engineering teams run reliable AWS and Kubernetes platforms, and put guardrails in place before AI agents get write access to production.

Book a 30-minute intro call See what I do
change-gate AI-authored change
Agent opened a merge requestterraform plan attached, no direct access
Plan classified HIGH RISKIAM policy change, database replacement
Policy as codeall rules passed
Human approvalwaiting for the platform owner
Applyonly from the approved plan
accountability follows approval, not authorship
10+ years in DevOps and SRE since 2015: SaaS, fintech, online classifieds and a software consultancy
4 teams led from a 4-person DevOps team to a 13-person platform and DevOps organisation; consulting independently since 2016
AWS · EKS Terraform and Terragrunt, GitLab CI, Argo CD, Datadog, Aurora and PostgreSQL
CET remote from Poland, on-site in Silesia when it helps

Services

Stable platforms first. Then AI agents that are safe to let in.

01

Infrastructure and reliability review

A fixed-scope look at your AWS and Kubernetes platform: account structure, Terraform, CI/CD, observability, disaster recovery and cost. You get the risks ranked by impact and a plan your team can act on.

written findings · risk-ranked roadmap · walkthrough with your team

02

SLOs, on-call and incidents

SLOs and error budgets that match what your users actually feel, alerts that page for the right reasons, runbooks, incident command and blameless postmortems that lead to real fixes.

SLO set · alerting policy · runbooks · incident process

03

Guardrails for AI agents in infrastructure

Before agents get write access: layered controls so every AI-authored change is classified by risk, checked against your policies, approved by the right person and applied exactly as reviewed.

risk model · policy as code · approval flow · audit trail

04

Hands-on platform work

When you need it built, not just advised: Terraform modules and standards, EKS upgrades, database major-version upgrades, GitLab runners on Kubernetes, migrations. As a short engagement or as an interim platform lead.

Talk about your project

Approach

How an AI-authored change gets to production safely

Guardrails, not bans. Agents keep their speed on low-risk changes, and people stay in charge of the ones that matter.

  1. 1

    Agent proposes

    Changes arrive as a merge request with a plan, never as direct access.

  2. 2

    Plan is classified

    Each plan is scored by risk: destroys, IAM, networking, data stores.

  3. 3

    Policy as code

    OPA and Conftest rules your team can read, extend and own.

  4. 4

    Human approval

    Low-risk changes flow; high-risk ones wait for the right approver.

  5. 5

    Apply as reviewed

    Apply runs only from the approved plan, with state protected.

  6. 6

    Watch for drift

    Drift detection and an audit trail close the loop.

Backstop Underneath it all: IAM and SCP limits that hold even when everything above fails.

Selected projects

Work I've delivered in production

Selected from more than ten years of production work, delivered as an architect, a team lead and a hands-on engineer.

AWS · EKS · Terraform

Multi-account AWS platform

Restructured AWS into separate accounts with dev, test and prod parity, running on EKS and managed with Terraform and Terragrunt.

Aurora MySQL · PostgreSQL

Database major-version upgrades

Ran the Aurora MySQL upgrade from 5.6 to 5.7 and oversaw the move to 8.0, rehearsed on production clones; did the same for PostgreSQL.

Datadog · SLOs

SLOs during an observability migration

Introduced SLOs and error budgets as part of a company-wide move to Datadog, so alerts reflect what users actually feel.

On-call · incidents

Incident management

Incident commander for critical incidents; wrote the alerting policy and runbooks and owned the postmortems.

GitLab · Kubernetes

Self-hosted GitLab on Kubernetes

Moved GitLab from Docker Compose to Kubernetes, with runners on the Kubernetes executor and shared pipeline libraries for teams.

Change management · fintech

Safe change in a fast-growing fintech

Led a DevOps team at a fraud-prevention SaaS and introduced a paced change-management process for infrastructure as the company scaled.

Kubernetes · consulting

Kubernetes for consultancy clients

Designed, deployed and supported Kubernetes platforms for software-house clients, and gave sales the infrastructure cost estimates behind the bids.

AI agents · policy as code

Guardrails for AI-authored changes

Terraform gates and risk scoring that let AI coding agents propose infrastructure changes without putting production at risk.

Working together

Simple to start, clear scope from day one

step 1

Intro call

30 minutes, free. You tell me where it hurts; I tell you honestly whether I'm the right person to help.

step 2

Fixed-scope assessment

Scope, timeline and price agreed up front. You get written findings and a plan, whether or not we continue.

step 3

Build or advise

Hands-on delivery with your team, or a monthly advisory retainer while your engineers do the work.

B2B invoicing through my company, BlackWebWolf. Remote from Poland in CET, on-site in Silesia.

About me

I'm a platform and DevOps architect based near Katowice, Poland, with more than ten years in DevOps and SRE. I've run infrastructure for a US SaaS platform, a fraud-prevention fintech and a global online classifieds business, led DevOps teams of 4 to 13 engineers, and consulted for software-house clients and independently since 2016. Most recently I spent five years as a platform architect and director, still hands-on in Terraform, CI/CD and incidents.

Now I focus on two things: keeping infrastructure stable, and making AI agents safe to use in it. I work in Python and Go, and I write things down: ADRs, runbooks and plans people can actually follow.

Available from January 2027

Let's talk about your platform.

Consulting engagements, interim platform leadership, or a full-time Staff role: tell me what you need.

Message me on LinkedIn michal@kajszczak.com
© 2026 Michał Kajszczak · BlackWebWolf Katowice, Poland · CET