Abstract

When given the opportunity to build a software engineering organisation from scratch, our vision was relatively solid — our structure solved many pain points we knew from earlier experiments. We had found a good balance between the people we could sustainably hire, the budget, and the products our research colleagues in a life science company needed. The thing we didn't have was a sustainable answer to: what do we do when we need to create more products and can't grow the staff linearly?

We had worked with external partners in those situations before, but we kept thinking we could keep strategic products in better shape if we didn't have to hand them off — if we kept them in the teams that built them. SRE came up internally as the answer to the hardest part of that — operations eating the team. We framed the proposal as an experiment, and that framing got our product peers' buy-in. This talk is told from two sides of that experiment: the manager who proposed it and built the engineering organization, and the engineer who became the first SRE and shaped what the practice looked like inside the product teams. Teams were designed for independence — kept apart so each could move at its own speed. The surprise after three years: in an organisation built for vertical independence, the SREs became the horizontal connection — carrying not just reliability work but communication, synergies, even capacity relief between teams.

Bio

Katharina Burckhardt is a platform and reliability engineer based in Cologne with 8 years experience across cloud engineering, infrastructure-as-code, and Site Reliability Engineering. Most recently in life science research IT, she designed and embedded an SRE practice across product teams, built reusable infrastructure modules, and led training programs on cloud and reliability topics. She works on the practitioner side of SRE — embedding reliability culture into product teams and making SRE principles a daily practice rather than a separate discipline.

Thomas Leyer is an engineering leader based in Cologne with 25 years in IT spanning systems engineering, scientific computing, DevOps, and engineering leadership. Most recently in life science research IT, he built a platform engineering team from the ground up and introduced Site Reliability Engineering as a function co-owning production with product teams. He works at the intersection of engineering practice and organizational design, with a particular interest in how organisations enable experimentation at every level — code, product, and the shape of the organisation itself.

Abstract

Most enterprise data landscapes grow organically into chaos: no lineage, no ownership, no data quality and AI can't drink from a swamp.

This talk shows how to build a central data lake that actually scales. Drawing on years of consulting work this talk covers the architecture, the engineering discipline (data mesh, Terraform modules, dbt tests) and, above all, the UI that decides whether you get a governed data lake or a swamp.

This talk then shows why a central lake is the cheapest path to safe enterprise-wide AI and how AI in turn lifts the metadata quality on which everything depends.

Concrete reference architecture, real engineering patterns, and the lessons we wish we'd had on day one.

Bio

Dr. Marc Schuh is a Principal Consultant at TNG Technology Consulting in Munich. He specializes in software and enterprise architecture, where Conway’s Law frequently comes into play and reveals that many technical problems are, at their core, organizational challenges. His approach focuses on driving steady improvement through intrinsic motivation and pragmatic solutions.

Marc served as one of the tech leads in building Europe’s largest commercial data lake, migrated a legacy Visual Basic application to C# using the Strangler Pattern, and supports large enterprises in improving complex marketing data pipelines. Beyond his consulting work, he contributes to innovative prototypes — including automated performance analytics for Germany’s national canoeing team and experiments with brain–computer interfaces.

In his spare time, he maintains open-source projects promoting digital independence from major cloud providers (github.com/MarcSchuh). Before joining TNG, Marc earned a PhD in physics and competed as a 400 m wheelchair sprinter at three Paralympic Games, becoming a world champion and European record holder.

Abstract

When was the last time you dealt with bare metal hardware? That dusty machine that has not yet been migrated to the cloud will always receive special treatment. If it crashes or a service fails, all hell breaks loose.

In this talk I'll give you a quick overview of Metal as a Service (MAAS). MAAS's dedicated goal is to consume bare metal machines the same way you consume VMs from a cloud. This also requires a mindset shift: treat machines like cattle with reproducible deployments, not like pets. I'll show you how this can be done in a homelab and how it looks at scale when running hundreds of machines across multiple racks.

Finally I'll spend a few minutes on how dogfooding and mandating SRE practices advanced our engineering team, and how we try to balance improving our own products through dogfooding while staying unblocked to innovate.

Bio

Thorsten is an Engineering Manager at Canonical leading distributed international teams of back-end and front-end developers in Canonical's Metal as a Service (MAAS) team. Before that he co-founded sustainabill (now Verso Supply Chain Hub), a cloud platform for supply chain transparency and traceability for physical goods. He did his PhD at the intersection of software engineering, requirements engineering, and machine learning, all these disciplines remain as important for his day to day work today as they were 10 years ago.

Abstract

Every user-facing flow in a microservices platform depends on dozens of downstream services. When teams only test the happy path, a single dependency failure can cascade into a major outage — and we've lived through that at trivago. Knowing the problem isn't enough. SRE and platform teams need a safe way for developers to inject failures before production does it for them — without breaking shared environments or paging the whole on-call rotation. In this talk we'll share how we built Emulate Fault: a developer-facing fault injection tool powered by Istio, our service mesh. We'll walk through the architecture — how we deliver personal, isolated failures (timeouts, errors, latency) scoped to a single session — and what resilience patterns on the application side actually make a difference when dependencies misbehave. Using a real incident and bugs we caught with this approach, you'll leave with practical ideas for bringing controlled, mesh-backed fault injection to your own platform — before a 3 AM pager proves the gaps for you.

Bio

Armin Aminian is an SRE at trivago, where he helps teams build and operate resilient systems across a large microservices platform serving millions of users.

Abstract

AI has dramatically reduced the cost of creating software. Today, agents can generate features, infrastructure changes, tests, and pull requests in minutes. What AI has not reduced is the cost of proving that software actually works. As software creation accelerates, validation becomes the new bottleneck. Every AI-generated change creates validation debt that must be paid before we can trust it in production. At Checkly, we began treating validation as a first-class artifact. Instead of generating only code, our agents generate executable specifications: unit tests, integration tests, synthetic monitoring, and production validation journeys. In this talk, we’ll explore how we use AI to generate testing and monitoring from the same understanding of a system, how this helps bridge the validation gap between development and production, and how we took our validation to the next level.

Bio

Daniel Paulus is SVP of Product and Engineering at Checkly, the synthetic monitoring platform built for any scale—trusted by teams from Indie Hackers to LinkedIn and Citi. He leads high-performing engineering teams, scales resilient systems, and stays hands-on with debugging and developer experience. Based near Berlin with his family, Daniel enjoys sharing practical lessons on shipping fast without breaking trust.

Abstract

Slow CI/CD pipelines delay builds, releases and deployments. A lot of time is wasted waiting for external dependencies like Docker images, Helm Charts, Linux packages, NPM packages, Maven packages, Go modules, Python libraries, and any other artifact that is required in the software development process. The registries that host these artifacts become a bottleneck. Besides performance and scalability concerns, organizations also struggle with cost, resilience and security risks when operating artifact registries. In this session, Thijs will show how to regain control over your software supply chain by caching your artifacts in Varnish and by using a Virtual Registry like Varnish Orca. Orca is a specially packaged version of Varnish that abstracts direct access to artifact registries. Orca accelerates access to artifacts by caching them as close to the consumers as possible, and uses its role as a “reverse caching HTTP proxy” to enforce security policies at request time, eliminate registry rate limits, protect pipelines against registry outages, and provide meaningful insights through an OTEL observability layer.

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

At enterprise scale, GenAI applications multiply the surface area of security and governance: across engineering teams, policy fragments into parallel implementations, spread across both the application platform and the data platform. A centralised DevOps platform consolidates that exposure into a single foundation: knowledge, security, and governance codified once into reusable CI/CD, IaC, and policy-as-code. A small central platform team then scales this across the organisation through blueprints and best practices.

Bio

Pavel Kovtun is a Solutions & Platform Architect at MobiLab Solutions in Cologne, with 19 years in IT across technology, product, data, and engineering leadership. He has recently led platform architecture at multiple multi-billion-euro enterprises, including DevOps platforms for GenAI workloads under regulated compliance.

Abstract

The SRE industry has built world-class system reliability. DORA metrics, SLOs, error budgets, automated recovery, the tooling layer has never been more sophisticated. Yet major incidents keep revealing the same pattern: technically excellent teams failing not because the system broke, but because the humans around the system couldn't coordinate effectively under pressure.

This talk introduces a framework for the layer most engineering organisations have never deliberately built: human reliability. Drawing on direct experience building cloud communities in Germany and Cameroon, and grounded in DORA research and Crew Resource Management principles from aviation, the talk proposes a 6th reliability metric, Incident Response Reliability (IRR), and makes the case for treating human coordination under incident conditions as a hard engineering discipline, not a soft skill.

Through a concrete case study of a Berlin fintech SRE team with elite DORA scores and a critical human reliability gap, the audience sees exactly what IRR reveals that DORA cannot, and how six months of tracking it dropped their MTTR from 29 minutes to 11 minutes without a single system change.

Bio

Endah Bongo-Awah is a Cloud Platform Engineer at Bedrock Germany, AWS Community Builder (Year 3), and Google Women Techmakers Ambassador. She is the founder of MMA Academy, a mentorship programme focused on career transitions into tech. She has spoken at All Day DevOps, AWS Community Day DACH, AWS User Group in Bonn and Cologne, and keynoted IWD 2026 at the University of Buea, Cameroon. She transitioned into cloud engineering in 2021, bringing over a decade of customer-facing and business leadership experience, a background that gives her a rare dual lens on both the technical and human sides of engineering reliability.

Abstract

Software systems today are still fundamentally reactive. Engineers wait for production incidents, investigate logs and traces, identify root causes, write fixes, review pull requests, deploy patches, and repeat the cycle endlessly.

But what happens when software can repair itself?

In this talk, Robert Markovski, Founder & CEO of Prilog, explores the emerging shift toward self-healing software systems powered by AI and autonomous remediation workflows. We’ll look at how modern observability, runtime telemetry, and AI agents can work together to automatically detect production issues, investigate failures, generate code fixes, validate changes, and propose pull requests with human-in-the-loop approval.

The session will cover: • The evolution from monitoring to autonomous remediation • Why traditional observability still leaves engineers overwhelmed • Real-world architecture patterns for self-healing systems • Safety layers, validation, and trust challenges in AI-generated fixes • The long-term vision of a world where software bugs continuously repair themselves before users ever notice

This talk is aimed at SREs, platform engineers, DevOps teams, and engineering leaders interested in the future of reliability engineering and AI-native infrastructure.

Bio

Robert Markovski is the Founder and CEO of Prilog, a startup focused on self healing software systems and AI driven engineering automation. Based in Cologne, he has spent years building full stack platforms, cloud infrastructure, and remote engineering teams across Europe and the US. Alongside leading Prilog, Robert also founded mozok GmbH, where he worked on large scale B2B software systems, platform engineering, and AI related products.

Abstract

When Fraunhofer invested in NVIDIA DGX systems to provide self-hosted AI inferencing to all Fraunhofer employees, the path from hardware to platform was full of pain points. This talk walks through our migration from NVIDIA's Base Command Manager to a Run:ai-based setup integrated into our existing Rancher-based Kubernetes infrastructure, covering the architecture on both sides: a custom stack handling GPU partitioning and inferencing through our own vLLM Helm charts before, and the Run:ai control plane after. We'll wrap up with a look at how the platform is put to use across Fraunhofer, from the FhGenie chatbot to a growing set of internal AI-powered applications.

Bio

DevOps Engineer at Fraunhofer specializing in Kubernetes and private cloud infrastructure. His daily work spans IaC, automation, and networking to build and operate the on-premises platform that now powers Fraunhofer's internal AI inferencing services.

Abstract

With the boom of Agentic AI naturally comes a question - what can it do for infrastructure, be it cloud or on-premises solutions? And what exactly should its role be limited to - using Agentic AI tools for deployment and documentation solely, or can it also take on the role of an operator? During our latest development, In ot teams, we've experimented with the possibility of using Agentic AI for infrastructure operations and would like to share our learnings, achievements, and thoughts for the future.

Bio

Infrastructure Project Manager with 7+ years of experience in IT

Abstract

Google trusts pods running in GKE. Every pod gets a native cloud identity, automatic credential roataion and seamless access to every GCP service. This is because they are running in Google infrastructure. Your onprem pods do not get any of that because they are running outside of GCP. They are strangers to Google Infrastructure. And the answer to this problem has been using service account keys, base64 them into a secret and hope no one takes a look closely. This talk talks about ending the pattern. Without moving your workloads to GKE and making Google extend the same trust to your on-prem Kubernetes clusters that it gives to its own. With Workload Identity Federation and OIDC, we built a platform at trivago where Google trusts every pod that runs on-prem and thinks it's running in it's own infrastructure. Allowing native support to access GCP Services, no key rotation, no tickets. No secret keys. We will cover this architecture end to end, and along with this an end to end demo.

Bio

Platform and Site Reliability Engineer at trivago

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.

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

Ravin Trivedi is a Senior Customer Architect at Imply with over 15 years of experience in the data space, specialising in real-time analytics, distributed systems and observability. He works closely with organisations across different industries to design scalable, performant and cost-efficient data architectures. With a strong customer focus, Ravin helps break down complex technical challenges and supports data/observability teams in turning them into practical, future-ready solutions that deliver measurable business value.