This year, the 6th edition of our TechSummit event took place in Pakhuis De Zwijger, Amsterdam. Once again, we hosted a range of passionate speakers to share insights on the latest AI and cloud infrastructure developments.
TL;DR
In this article, Robert van der Meulen, Director of Product Strategy at Leaseweb, provides an overview of the TechSummit Amsterdam 2026 event, analyzing the themes explored by the speakers. For Felix Meyer from Cloudeteer, observability was the focus. Paul Farver and Lorenzo Setale from The LEGO Group walked us through their experience with robust constructions. Both talks had the theme of resilience at the heart of their message.
Resilience – do you know what’s going on in your environment?
As Director of Product Strategy at Leaseweb, I’m proud of the work that has gone into organizing TechSummit. We fill a need that is rare to find these days: a conference focused on content that inspires a technical audience, without the “fluff” typical of more commercially oriented events.
This year, topics focused on the theme of ‘building resilience at scale’, which has a lot to do with knowing what is going on in your environment.
What can you do when results don’t meet expectations?
Platforms and applications are becoming increasingly complex. The use of all kinds of abstraction layers, like virtualization, containerization, and various frameworks add to the complexity. Getting valuable insight into what your platform is doing can get complicated quickly. Combined with the need for continuous rolling deployment changes, it can be easy to lose track of what’s happening, especially when things start to break.
Felix Meyer, Chapter Lead of Operations Engineering at CLOUDETEER, presented a talk about observability in systems such as these.
Because of the complexity in modern systems, even while everything is ‘running fine’, the business can still be unsatisfied by the results. The interesting direction that Felix chose was to consider three different angles when observing if a system is functioning. A stack can be running fine, but the results can still be “slow, expensive, or wrong”.
As this was an engineering talk, of course Felix had a solution to define metrics. He also measured them (in this case with OpenTelemetry) and monitored them so action can be taken.
The talk also covered the concept of measuring AI inference speed, cost, and accuracy. As it’s considerably more difficult to monitor GPUs for performance and use, seeing how CLOUDETEER solved this in production was insightful. Another learning point highlighted is to label your traces, so you can track errors across the stack and pinpoint the problem.
If you would like to learn more, Felix’s sources and examples of the implementation are available on GitHub.
How do your systems respond to failure?
The resilience of an environment can also be reviewed by measuring how systems respond to failure. Paul Farver and Lorenzo Setale, Software Engineers at The LEGO Group, presented a talk on the creation of a ‘Kubernetes as a service’ platform for their organization.
There were many teams working on and deploying to this platform. This meant that as well as the technological aspect of making things work, a culture change in terms of how the various teams interfaced with the Kubernetes service also occurred.
Users – in this case, engineers and developers – had to adopt a GitOps way of working to deploy to the platform. Part of that change was also to make sure that apps follow cloud-native standards, don’t persist data (we’re using GitOps, after all!) and can sustain failure.
Unfortunately (but also fortunately, thankfully), failure doesn’t happen all of the time, so the talk took an interesting twist:
Why not introduce failure into the stack systematically?
It was explained that this approach can result in a way of working where parts of the stack are artificially ‘failed’ to make sure that applications and deployments know how to deal with them. This includes ensuring core principles and architectural standards are followed.
To the LEGO Group, resilience is not simply a matter of documenting architectural standards. Paul and Lorenzo’s platform team deliberately introduces interruptions into Kubernetes workloads, including through descheduling and node lifecycle management. This exposes applications that cannot tolerate disruption and encourages developers to build safeguards before production. Their experience also highlights a broader lesson: architectural policies must be enforced, not merely documented. Otherwise, something as routine as a Kubernetes upgrade can become a complicated exercise involving multiple application teams.
Join us in Amsterdam for TechSummit 2027
Unfortunately, we could not record Paul and Lorenzo’s talk, so to avoid future disappointment, be sure to attend TechSummit 2027!
We’re currently working hard on the planning, with announcements about tickets and location to follow soon.