Talks at conferences

Cloud promises elasticicy and cost savings, but the reality of most enterprise organizations is different: overflowing budgets, unreadable dashboards and developers who copy-paste resource configurations from StackOverflow or ChatGPT. FinOps is not just about cutting costs, it’s about culture of shared responsibility of your cloud spendings. In this talk, I’ve shown a practical guide, how to embed FinOps principles into your organization on top of a Kubernetes platform. From resource configuration, Cluster Autoscaler and Karpenter configuration, and chareback model. No theory, no vendor demo. Just architecture, which works in production.

I’ve talked about how to build a zero-trust solution in the Kubernetes world, using the Cilium CNI, mTLS and how to utilize Hubble to collect network events. And all of that nicely pack into reports for example for NIS2 directive.

This was my first time speaking at Installfest and I did not hold back. I gave a talk about how AI agents (primarly Claude Code) take care of Kubernetes cluster for me. I shared my experience, what worked and what did not. There are three main takeaways:

• Declarative configuration – agent can read all configuration for its context
• CLI tools – agents can investigate and validate the real state on it’s own
• Objective validation – agents need tools (CLIs, MCPs, see previous point) to verify, you also need to provide the verification criteria in your prompt

A recording of my talk is available on YouTube, but in Czech.

I spoke about ways to secure your clusters, trust applications, monitor for data leaks, and defend against supply chain attacks.

I covered topics ranging from basic Kubernetes primitives like securityContext and NetworkPolicy, through image signature verification, all the way to runtime-enforcement engines such as Falco, Tetragon, and KubeArmor.

At KCD, after a year (this time in Bratislava), I gave a talk on attestations and OCI image signatures — how to use them and how to validate image signatures and whitelist registries from which images in Kubernetes are allowed to run. Slides (PowerPoint presentation) are available for download here.

If you want to try it yourself, check out the demo project on GitHub, the recording on YouTube, or get started on your own with Sigstore and Cosign!

At this year’s KCD in Bratislava, I also participated in a panel discussion on the topic of AI/ML Meets Cloud Native. I shared some of my insights and experiences, for example from GLAMI, where I worked on MLOps and operated recommendation systems serving 14 countries where GLAMI was active at the time. A recording is available on YouTube.

At P2D2, I gave an updated version of my talk from KCD Prague 2024, with a few adjustments for an audience less familiar with Kubernetes. The talk was about running a highly available Postgres cluster on Kubernetes using CloudNativePG — a Kubernetes operator.

At Kubernetes Community Days Prague 2024, I gave a talk on running a PostgreSQL cluster on Kubernetes. How to do it properly and sleep well at night without PagerDuty waking you up, using the CloudNativePG Kubernetes Operator.

I wrote a short blog post about KCD.

A talk on self-hosted vs managed solutions. What are the differences, what are the individual advantages and risks, and how to actually choose between them.

A short talk with a live demo on how you can scale on Kubernetes beyond the classic HPA approach. KEDA is a great tool that helps you scale using a wide range of so-called Scalers. For example, based on messages in a RabbitMQ or Kafka queue, events in AWS Event Hub, or simply using a cron schedule (eight pods at 8 AM, only three pods at night), and many more. Today, KEDA offers over sixty scalers plus the option to write your own.