Abstract

LLMs are confident liars. We will start with a true story about an internet meme, and an AI that learned exactly the wrong lesson, and use it to explain why hallucinations happen and how to stop them This talk walks through a production-ready approach to grounding LLMs in real data using Retrieval-Augmented Generation on OpenSearch. We will compare RAG against fine-tuning and explain why retrieval wins for fact-based use cases. We will dig into hybrid search, combining BM25 lexical scoring with k-NN vector search, and show how to tune the normalization and weighting that ties them together. We'll spend real time on chunking strategies because chunking is where most RAG projects quietly fail. Finally, we will introduce a three-layer evaluation framework covering retrieval quality, configuration, and answer faithfulness. You will leave with the mental model that turns RAG from guesswork into engineering.

Bio

Senior Platform Engineer at enmacc, OpenSearch Ambassador, and PyLadies Munich leader. I spent years as a backend developer, and now I build platforms so other developers do not have to suffer like I did. Multi-cloud by trade, search nerd by choice, continuous learner by habit.

Helping scale Europe‘s leading OTC energy trading platform and driving more transparent, efficient markets for the energy transition.

Abstract

Production is a wilderness. The terrain shifts every minute, meaning human and agent failures are not caused by missing intelligence. They come from acting on stale context. Discovery takes time, and in production it is often the most expensive part of the workflow.

In this talk, I show why AI SRE agents fail when they rely on runtime discovery, and what changes when they operate on live context instead of snapshots.

The short version: an AI SRE that admits uncertainty is the only kind you can safely let run.

Bio

Co-founder and CEO of ewake, Poone Mokari builds AI SRE agents grounded in a live system map. Prior to founding ewake, she worked as a Site Reliability Engineer at companies including Criteo, where she gained hands-on experience managing large-scale production systems and incident response.

Abstract

More and more organizations are realizing that observability is the missing piece in how they deliver and operate digital products: teams get pulled into firefighting, troubleshooting stays siloed, and leaders lack a clear line from system behavior to business outcomes. In this talk, we share first-hand experience designing an organization-wide observability rollout in a complex environment—guided by SRE principles and treating reliability as a product. Rather than starting with a mass, technology-first onboarding, we start from business objectives and a clear value proposition. These are then translated into clear requirements, which are used to select an observability platform that matches your custom need, scale, and constraints. We then show how to define the right level of onboarding based on your product portfolio, and how to align early with product and delivery stakeholders so expectations, ownership, and priorities are clear from day one. You’ll learn a practical approach to structuring service onboarding across business, application, and infrastructure layers to achieve end-to-end visibility and correlation across services and domains—where the most valuable insights emerge. Finally, we cover how to enable engineering teams early and set up an operating model with shared goals, shared responsibility, and dedicated roles to drive adoption at scale. Our goal is to provide attendees a repeatable blueprint to design an observability launch that fits their organization.

Bio

Stefan Lenz is an experienced IT Solution Architect at Siemens with nearly three decades of experience in designing and delivering enterprise technology solutions. He specializes in cloud computing, virtualization, and IT-driven services, with a strong focus on helping organizations achieve business outcomes through scalable, efficient architectures. With a background spanning project management, SaaS, and service processes, Stefan combines deep technical expertise with strategic insight to drive digital transformation initiatives.

Michele Dodic is an Observability Manager at Siemens Mobility, leading an enterprise-wide observability rollout. As a solution architect and active member of several SRE and DevOps communities, he promotes SRE principles, practices, and culture to help teams improve reliability and operational efficiency. Michele regularly shares his insights as a public speaker at conferences including Splunk .conf, Dynatrace Perform, All Day DevOps, ChaosCarnival, Conf42, and other industry events.

Abstract

So you finally got your organization to invest in OpenTelemetry. You carefully evaluated observability backends and picked the perfect one. Everything is awesome. Then twelve months later, your costs have skyrocketed and you can’t explain why. What happened? This talk examines how to emit meaningful telemetry while keeping costs under control, by exploring the following: - What to actually instrument - Which metrics to focus on - Pipeline efficiency with OTel Arrow Applying sampling, filtering, and intentional instrumentation to cut down on noise Schema management and validation with tools like Weaver We’ll review the ingredients of a mature observability implementation with OpenTelemetry: one that grows with you instead of overwhelming you. You’ll learn how to apply cost-effective techniques to achieve meaningful observability. Speaker Notes (visible to organizers only) As more organizations embrace OpenTelemetry and mature their Observability practice, we find ourselves coming out of that Observability and OpenTelemetry "honeymoon period". We've gone from, “We’re using OpenTelemetry, therefore we have Observability” to "How do we actually make this work for us?" This talk will equip organizations to build a sustainable and long-lasting observability practice built on OpenTelemetry.

Bio

Josh Lee is an Open Source Advocate at Altinity, focused on ClickHouse, Kubernetes, and OpenTelemetry. With more than 20 years of software engineering experience, he has worked across developer advocacy, product management, observability, and full stack development. Josh is also an organizer of the Open Source Analytics Conference and a frequent speaker in the cloud native and observability community.

Abstract

Imagine an Oncall shift where you don’t start your day sifting through routine capacity alerts, nudging stuck rollouts, or closing noisy, low-impact tickets. Instead, you get to tackle things that matter. This is the goal we’re chasing at Google.

We are developing a system where software agents can autonomously handle a significant chunk of operational toil. The key is to do this generically and horizontally, making the solutions broadly applicable crossing the lines between developers and operations.

In this session, I’ll share our journey and lessons learned. We’ll cover the significant challenges, including evaluation, ensuring safe and secure operations, and how to codify complex, sometimes opinionated, remediation steps. I’ll outline the infrastructure we’ve put in place due to those challenges and requirements.

This talk aims to provide a practical perspective on leveraging automation and agents in a production environment. You’ll leave with critical questions to consider for your own agent that interacts with production.

Bio

Chiara Dellantonio is a Munich-based Senior Site Reliability Engineer at Google. Over her seven-year career, she has served as a Tech Lead across various teams supporting internal products and the authentication stack and developing production monitoring infrastructure. She currently focuses on applying intelligent automation and AI capabilities to production operations, making system management safer and more efficient. Her passion for collaborative problem-solving extends beyond systems engineering: Chiara is an elite speed jigsaw puzzler, boasting a World Team Championship, a European Championship, and two World Pairs silver medals.

Abstract

After five years of managing serverless databases, I have learned that my rollercoaster journey is very similar to the CPU usage you dream of seeing in the console. This session shares five hard-earned lessons learned while working with so-called serverless databases.

Bio

Renato has extensive experience as a cloud architect, tech lead, and cloud services specialist. He lives in Berlin and works remotely as a principal cloud architect. His primary areas of interest include cloud services and relational databases. He is an editor at InfoQ and a recognized AWS Data Hero.

Abstract

Every week, one engineer on my SRE team takes the gatekeeping shift. They handle everything that arrives — every interrupt, every ad-hoc request, every ticket without context or priority. Everyone else gets to think. It is, as we still say internally, a horrible week. Triage under pressure is not a new idea. Emergency rooms do it. ITIL codified it. Kanban made it visual. The principle is the same: not everything that arrives is equal, and treating it as equal is how teams drown. We did the same thing for SRE. We got very good at it. That was the first problem. By the time the gatekeeper role was mature, it was fast, accurate, and resolving things in hours that used to take days. And then we stepped back and looked at what it was resolving. A significant portion of the work flowing through that seat shouldn't have been fixed by SRE at all. It should have been automated. Or it should have been rejected as a requirements problem — something that needed to be addressed upstream, not patched downstream on a weekly basis. We had built a perfect team for doing the wrong things quickly. The role shifted. Not triage-and-resolve. Triage-and-surface. The gatekeeper's job became identifying where the real problem lived — automation candidate, requirements gap, architecture debt — and routing accordingly rather than absorbing it. TOIL peaked at 83.9% in 2023 during three simultaneous acquisitions. By 2025 it was 44.7%, inside Google SRE's benchmark. The gatekeeping rotation was the practice that moved that number — not by going faster, but by going to the right place. Rotating who sits there is what stops it becoming one person's problem. Engineers who've done the horrible week know exactly which ticket types need to stop existing. Practices cross-pollinate. Stakeholders get exposed to different people asking different questions. The P&L context shifts as ownership spreads. In 30 minutes: how we implemented it, how we almost ruined it by optimizing the wrong thing, and what the data looked like when we corrected course. You can start this next sprint. No budget, no organizational permission required — just one week, one person, and the willingness to ask why it keeps coming back.

Bio

Andrea Valenti is Senior Director of SRE leading 38 engineers across multiple geographies. Over six years, three mergers and a couple of divestments, he developed a framework for building engineering organizations that absorb disruption rather than accumulate it. Before technology, he worked as an educator in psychiatric care in Bologna — which taught him more about teams under pressure than any management book. He plays saxophone badly, builds electronic instruments in his spare time, and developed most of this framework between Brussels, Miami, and Mexico City before settling in Barcelona.

Abstract

AI can already help write and test code, but many teams still do not want fully automatic changes. In this talk, I will show a simple Human-in-the-Loop Agentic SDLC setup where AI agents can plan work, update code, and run tests, but important actions still need human approval. I will show the MVP demo and also compare a few free alternatives for the coding layer, including Cline, Continue, and OpenHands.

Bio

Boris Zaikin is a Lead Cloud and Software Architect and technical leader specializing in Azure, AWS, and GCP, with a strong focus on data and AI. Based in Munich, he has a proven track record of designing and delivering scalable cloud and platform solutions, and is also an active tech speaker and blogger.

Abstract

With all the hype around OpenTelemetry, you may be tempted to use OpenTelemetry and its SDKs for all of your application instrumentation needs. However, when it comes to generating metrics for usage in Prometheus, you should at least think twice before going all in on OTel. Not only do you risk throwing away some of the core features that define Prometheus as a monitoring system, but you'll also end up with awkward metrics translation and escaping issues, as well as other inefficiencies and complexities. That's why I still recommend using Prometheus's own native instrumentation client libraries over the OTel SDKs if you want to get the best possible Prometheus monitoring experience. In this talk, I'm going to lay out some of the major reasons that led me to this opinion.

Bio

Julius co-founded the Prometheus monitoring system and originally led the project to success at SoundCloud and beyond. He now focuses on growing the Prometheus community, started PromCon, the first conference around Prometheus, and helps companies use and adapt Prometheus through his company PromLabs. Before creating Prometheus at SoundCloud, Julius was a Site Reliability Engineer at Google.

Abstract

Come discover how to code a Game Master that will orchestrate multiple AI agents, each specialized in a particular task. We'll explore concepts like agent tools, the MCP (Model Context Protocol), A2A (Agent to Agent), and RAG (Retrieval-augmented generation) to create agents and/or MCP servers that roll dice, verify DnD rules, and generate the story as it unfolds. Whether you're a role-playing game fan or not, there will be things to learn during this workshop :)

Bio

Helping Devs build cool stuff, learn faster, and laugh along the way — one bug at a time. 𝘈𝘭𝘭 𝘷𝘪𝘦𝘸𝘴 𝘦𝘹𝘱𝘳𝘦𝘴𝘴𝘦𝘥 𝘢𝘳𝘦 𝘮𝘺 𝘰𝘸𝘯.

Abstract

Grafana dashboards are often built when we don’t have time to build them and used during incidents, when we don’t have time to understand them. In this talk, we’ll dive into real-world examples and explore multiple levels of monitoring through screenshots and stories from the field. By the end of the session, you’ll walk away with a few simple yet powerful principles to design and organize Grafana dashboards that truly help during critical moments.

Bio

Software Architect with extensive experience in big data and high-load systems. Passionate about designing highly scalable distributed systems with strong emphasis on observability, monitoring and documentation.

Abstract

Running a Telegram bot as a single pod is easy. Scaling it to multiple replicas is where things get interesting. This talk follows the real evolution of a production bot for a language-learning platform. Each step was driven by a concrete problem: polling breaks when two instances race for the same update, so you switch to webhooks and introduce an ingress. In-memory sessions disappear between replicas, so Redis shows up. Reminder workers block the webhook handler, so they become separate binaries running as CronJobs. You add a second language pair and suddenly the same Helm chart needs to deploy two isolated bots — and a subtle bug surfaces where one bot answers for the other. We'll also cover how LLM calls made cost invisible (and how a small in-cluster usage-tracking service fixed that), plus encrypted secrets in git with SOPS across a multi-chart monorepo.

Bio

Vasil Kulakov is a Staff Engineer in Site Reliability Engineering with a strong focus on building and scaling production infrastructure. Currently at snappyloop.work, he has helped establish the backbone for an AI-driven language learning platform, supporting rapid user growth. Previously, he led SRE teams at CHECK24, combining hands-on engineering with leadership across high-performance systems. His work spans Go, PHP, and modern cloud-native tooling, with a practical, execution-first approach to reliability and scale.

Abstract

Public cloud providers offer convenience — but at the cost of control, transparency, and increasingly, budget. What if you could have the same self-service Kubernetes experience, on your own hardware, built entirely on open source technology? This talk walks through the full stack of building your own cloud platform to offer Kubernetes — layer by layer. We start at the bottom with Networking and Compute. From there, we move up to Metal-as-a-Service with metal-stack.io, which turns bare-metal servers into programmable, API-driven cloud infrastructure — with on-demand provisioning, physical tenant isolation, and no hypervisor overhead. On top of that, we explore how Gardener and ClusterAPI handle Kubernetes cluster orchestration at scale: self-service cluster provisioning, automated lifecycle management — delivering the public cloud feeling, without the public cloud dependency. We will also touch on what owning the complete stack means for SRE teams: from observability across every layer, to the ability to see exactly what happens during debugging — because there are no black boxes, only open source code you can read, trace, and understand.

Bio

Simon Mayer is working as a Cloud Native Engineer at x-cellent technologies GmbH. The company is focused on building sovereign cloud and Kubernetes platforms built entirely on open source technology — for organisations with the highest demands on security, compliance, and control.

Abstract

Reliability fails in surprising ways—not because engineers don’t care, but because our brains are optimized for speed, not accuracy. In high-pressure IT environments, that leads to familiar patterns: premature conclusions during incidents, “obvious” fixes that don’t fix anything, risky changes that felt safe at the time, and meetings where everyone nods while the real problem stays untouched. This lecture is a sharp, highly relatable look at the human side of reliability: how bias shapes decisions in operations, why smart teams repeat the same mistakes, and what lightweight practices help you improve outcomes without adding bureaucracy. Expect real-world scenarios, humor, and practical takeaways you can apply immediately—without me giving away the full playbook.

Bio

I’m Marcel Koert—a Site Reliability / DevOps / Platform Engineering professional focused on turning “reliability theory” into a way of working teams can actually sustain. I’m the author of two books in the Essential SRE series: • Essential SRE: Way of Working — a practical operating model for SRE: how to run reliability day to day, reduce chaos, and make delivery and operations work together (without hero culture). • Essential SRE Articles — a curated, straight-to-the-point collection of the essential knowledge SREs need, grounded in real-world application rather than slideware. Alongside the books, I share talks, playbooks, and field-tested guidance through MeloMar-IT to help engineers and leaders build systems (and teams) that stay stable under pressure. If you care about reliable delivery, calmer on-call, and reliability that survives contact with reality—let’s connect.

Abstract

Now that the Nginx Ingress Controller for Kubernetes is deprecated, a lot of organizations are re-evaluating how they are going to expose Kubernetes services to the outside world. Gateway API is the obvious successor of Ingress and in this presentation Thijs will introduce the brand new Varnish Gateway Controller for Kubernetes. Varnish is a well-known reverse caching proxy that accelerates any type of HTTP workload. The open source project has been around for more than 15 years and is often deployed inside Kubernetes clusters to accelerate individual services. In this presentation Thijs will show how the power of Varnish adds caching to Kubernetes at the Gateway API level, while still supporting all the routing capabilities that a Gateway Controller is known for.

Bio

As the Technical Evangelist at Varnish Software, Thijs Feryn focuses on web performance, software scalability, and content delivery. He demonstrates content-driven and technical messaging through presentations, videos, books, blog posts, social media posts, podcasts, and other media. Thijs is a published author and wrote Getting Started with Varnish Cache and Varnish 6 by Example. As a public speaker, he has a track record of over 380 presentations in 26 different countries, where he is often praised for his energetic and engaging presentation style. As an evangelist, Thijs is also active in many open-source communities, most notably the Varnish and PHP community. He has contributed to various communities for over 15 years both technically and as an organizer and facilitator. Prior to joining Varnish Software, Thijs Feryn spent 15 years in the web hosting industry, tackling web performance and scalability issues on a daily basis and evangelizing these topics. For more information about Thijs’ past & upcoming presentations, please visit https://feryn.eu/speaking.

Abstract

And practical info

Abstract

IT systems are not neat self-contained boxes with a few components. A single application deployment today has hundreds of moving parts. An IT landscape contains thousands of those applications.

On the other side sit AI agents with vast world knowledge. Like humans they struggle with context overload. They can't possibly keep a whole datacenter in their head at once.

Neither can we. We navigate.

In this talk we'll show how an agent can navigate and operate a system of hundreds of thousands of components, working from a live map of the infrastructure and a connector layer that lets it act on anything the map reveals.

The result: an agent that adapts itself to the landscape it walks into, instead of a bespoke setup for every shape of IT.

Bio

Benjamin Hofmann is CPO at Hyground, where he focuses on resolving software incidents in minutes and improving operational flow for engineering teams. With a background spanning Testing and Agile methodologies, DevOps, Cloud Native development, and Observability, he brings a cross functional perspective to the full software lifecycle. Based in Rosenheim, Germany, Benjamin is passionate about building impactful technology that enhances productivity, collaboration, and quality for both teams and end users.

Abstract

Digital sovereignty goes far beyond data protection and IT security. Europe’s digital future currently depends on a small number of global tech providers, bringing risks such as vendor lock-in and geopolitical exposure.

For platform engineering teams, these concerns quickly become practical decisions. Which workloads can run on which platforms? Which dependencies are acceptable? And how meaningful is sovereignty once you move past marketing claims?

This talk examines what digital sovereignty means in real-world platform engineering. It covers the key challenges, the trade-offs teams must navigate, and how organizations can build viable strategies. Drawing on hands-on project experience, it outlines pragmatic approaches to balancing technological capability with independence and compliance.

Bio

Michael Harrer is Co-Head of the Cloud Sovereignty and Compliance business unit at MaibornWolff. He previously led the DevOps and Cloud Native department for several years.

As a Principal Architect, he focuses on designing cloud platforms that combine technical excellence with responsibility and independence. His work helps organizations translate multi-cloud, data sovereignty, and compliance requirements into sustainable platform strategies.

Abstract

Abstract

On-call incidents don’t fail because teams lack dashboards. They fail because observability systems slow down under real investigative load

As telemetry volumes grow and retention windows expand, SRE teams are being asked to run deeper, broader investigations—often under time pressure—on platforms that were designed for steady-state monitoring, not bursty incident response. Tightly coupled observability stacks bind storage, compute, and query together, forcing teams to overprovision infrastructure, limit retention, or accept degraded performance during incidents.

In this talk, we’ll explore why decoupling observability architectures is becoming essential for SRE teams operating at scale. Using a real incident investigation workflow, we’ll break down how separating data storage, compute, and interaction layers allows teams to keep fast, reliable monitoring while elastically scaling investigations when incidents occur.

Bio

Kyle Hoondert is a Senior Customer Architect at Imply with over 15 years as an experienced solutions architect and Unix/storage administrator, designing, implementing and supporting infrastructure and storage solutions in managed service organisations.

Abstract

The next evolution of incident response isn’t faster alerts, it’s autonomous resolution. Join ilert CEO Birol Yildiz as he shows how AI SRE agents now diagnose and remediate outages without waking anyone up. Learn how these systems combine observability data, deployment context, and code intelligence to restore services in minutes and hand over clean incident reports instead of 3 a.m. pages.

Bio

Birol Yildiz is the Co-founder and CEO of ilert, adeptly steering the company with a rare combination of technical and product expertise. His prior experience includes a significant role as Chief Product Owner for Big Data products at REWE Digital. With a strong foundation in computer science, Birol bridges the gap between developer and product strategist, constantly striving to innovate and provide customer-centric solutions at ilert.