Open Source Summit Europe 2025

When we talk about cloud security, the first thing that comes to mind is otters. Obviously.

In this talk, we’ll dive into the fascinating analogy between the world of cloud security and wildlife. We’ll explore Falco, eBPF, and others and how they relate to our very own animal ecosystem. This perspective offers a unique and intuitive way to understand cloud security.

Think of chameleons and adaptive security: changing colours to match evolving threats.

We’ll migrate like birds, discovering security’s role in real-world workload transitions across clouds.

We’ll collaborate like bees, protecting the hive with policies, alerts, and shared responsibility.

We’ll defend like armadillos, layer by layer, showing how strong defence is often the best offence.

And we’ll feel our spider-sense tingle — like observability tools catching anomalies across the great web of services.

And, just like otters, we’ll hold hands to stay together in the rushing currents of the cloud.

And who’s heard of chaos monkeys? Clever and unpredictable, just like attackers exploiting misconfigurations and edge cases! We’ll take a look at how that plays out in the wild world of cloud-native security.

In this talk, we will present open-source tools for evaluating and securing AI models which are key in building responsible AI systems. We'll start with an overview of these tools and their use cases, organized through a simple ontology to help classify them.

We'll cover:

- Tools that assess bias and fairness, such as AIF360

- Evaluation tools like Garak, which offers comprehensive security and safety assessments for LLMs; Promptfoo, which supports prompt engineering and testing; and Giskard, which allows the use of custom evaluation datasets

- Guardrail tools for LLM systems, like NeMo Guardrails

- Tools that focus on prompt security and insights, such as LLMGuard and LangKit

- Tools for traditional ML model security, like the Adversarial Robustness Toolbox

The main goal of the talk is to give attendees a clear overview of the wide range of open-source tools available for securing AI models and provide examples and insights about them that can help in making an informed decision on which one to use.

This talk is about software developers' journey into AI Agent creation.


We'll explore the evolving toolkit, from specialized IDEs like Cursor and VS Code (with AI Agents extensions) to frameworks like LangChain and LlamaIndex that simplify complex workflows.


I'll share firsthand experiences comparing local development with models like Ollama, LM Studio and LLM Runner against production deployments using OpenAI and Anthropic APIs. We will explore how to consume MCP Servers to provide tools to our Agents.


We'll address critical decision points: when to use RAG versus fine-tuning, how testing differs from traditional software, and the unique debugging challenges of non-deterministic systems. Finally, we will look at the challenges of deploying AI Agents in production and what we can do as developers to minimise the risk.


Developers will leave with a clear understanding of the technical shifts required when building AI agents, available tools, and practical strategies for overcoming the most common obstacles in this rapidly evolving space.

Linux kernel code (and any C project in general) can greatly benefit from several static code analyzers. Let's dive into open and free (as in free beer as well) tools for static code analysis used in the Linux kernel, because there is more than one tool

The session will describe in detail how to check your Linux kernel code around build time to improve the code quality.

Many techniques will be applicable to other projects, but session focuses on the Linux kernel due to its own build process.

Helping newcomers step into the Linux kernel community is not an easy task, but

with the right mentorship it becomes a transformative journey for both the mentor and the

the mentee.


This talk shares lessons from over a decade of hands-on mentoring experience, helping students

grow from their first patch to becoming meaningful contributors.


This presentation focuses on real world examples from subsystems like IIO, ALSA and device tree

bindings and explains the practical steps to get beginners started. Each steps is important:

from selecting the appropriate tasks and hardware platfroms, to explaining kernel workflows

and mailing list culture.


Attendees will gain practical guidance on building effective mentorship practices, learning

how to support newcomers as they evolve from learners to contributors. Also, this presentation

is valuable for newcomers looking to take their first step into kernel development, offering

a clear view of the journey ahead and how to access mentoring programs like Outreachy Program for Women

or Google Summer of Code.

In a world where cloud-based AI solutions dominate the tech landscape, the power and flexibility of local AI are often overlooked.

This session will showcase how you can leverage open-source AI models to streamline your development workflow, improve data privacy, and reduce costs. We’ll explore tools like Ollama, LM Studio, and Open Web UI to replace Github Copilot and ChatGPT on your local machine.

Through hands-on demonstrations, you will learn how to integrate these tools into their daily tasks, making your development process more efficient and secure.

Whether you are a developer or tech enthusiasts, this session will provide practical tips you can apply directly.

Traditional software is built around transactions—predictable inputs and outputs. But agentic AI systems are different: they’re conversation-driven, adaptive, and capable of reasoning over time. In this talk, we’ll break down what makes AI systems "agentic" and the key components required to build them. You'll learn about the three essential stacks—training, inference, and runtime—and dive deep into what makes an agentic runtime work: durable execution, memory, agent lifecycle orchestration, and tool use. Finally, we’ll explore the open-source ecosystem powering agentic AI systems, including popular open source projects such as vLLM, LangGraph, and AutoGen, and explain where they fit in the agentic stack.

Intrigued by quantum computing but unsure where to begin? This beginner-friendly session charts a practical path using accessible, open-source tools. Discover how open-source Python libraries like Cirq and Qiskit makes defining your first quantum circuits for simulation surprisingly straightforward. We'll then explore how to take these Python-based simulations, package them into standard containers, and run them directly on Kubernetes – the ubiquitous platform for modern applications.

See a live demo running a basic containerized Python-based quantum simulation as a standard Kubernetes Job, showcasing deployment in this powerful environment.

We'll also map the road ahead: as needs scale, understand how the open Kubernetes ecosystem could support advanced management via CNCF tools (like Kueue) for K8s-native queueing and HPC scheduler integrations (like Slurm) for larger batch workloads.

Get a clear roadmap from your first Python script to running Quantum Simulations on K8s, understanding the path towards tackling larger challenges using open, cloud-native approaches.

Most Linux users use it every day: the bash shell. While Linux newbies only sporadically access the command line (terminal), their familiarity with it increases with continued use.

However, very few people take advantage of the sheer power of the bash. A basic understanding of the Bash grammar will help you get started: from simple commands, pipelines (|), lists, to compound commands, coprocesses, shell function definitions.

The Bash is incredibly versatile.

Quoting, for example, is used to remove the special meaning of certain characters or words to the shell. Values can be stored in parameters - of which there are several: positional parameters, special parameters, shell variables (e.g. $BASH, $LANG, $PATH, ...), arrays. Functions, arithmetic evaluation, conditional expressions and exit status are features often used in Bash scripts. Last but not least, shell bulletin commands are also part of the bash feature list.

Join Werner's talk and discover many possibilities for your favourite shell that you may have only dreamed of.

As the demand for intelligent sensing solutions increases, integrating real-time AI capabilities into edge devices can lead to development inefficiencies. The fragmentation of various IoT devices and software stacks adds to this challenge, requiring advanced embedded software development that considers resource limitations. At the same time, technologies like WebAssembly (Wasm) and image recognition AI models are becoming more accessible. To effectively leverage these innovations, we must expand the development ecosystem for edge sensing devices and simplify the development process.

In this session, we will explore strategies to provide a consistent and scalable approach across different hardware environments. We will introduce our open-source software suite that enables:
- Easily develop Wasm-based applications using our SDK
- Integrate seamlessly into edge AI camera devices
- Deploy Wasm applications on the edge with a graphical user interface

Join us to discover how our platform is changing the edge AI landscape and making intelligent sensing more accessible.

In this talk, Joshua will share his insights and experiences with OpenTelemetry, an open-source project that offers protocols, APIs, and SDKs for collecting metrics, traces, and logs from applications and services. He will cover the comprehensive toolkit provided by the OpenTelemetry community, including language SDKs, the Collector, and the OTLP formats for metrics, traces, and logs.

He will demonstrate how to instrument and monitor a microservices application running on a Kubernetes cluster, utilizing the full potential of OpenTelemetry. Attendees will learn how to use powerful open-source tools like Jaeger and Prometheus to effectively analyze telemetry signals from their applications.

By the end of this session, attendees will have a solid understanding of how to implement OpenTelemetry in their projects, enhancing their debugging and observability practices. Join us as we delve into the world of OpenTelemetry, unlocking the capabilities of this powerful technology for your development needs.

As foundation models become increasingly central to health applications, the open-source community plays a vital role, not only in democratizing access to data and models, but also in driving safe, reproducible, and equitable innovation. In this talk, I will explore the rapidly growing ecosystem of open-source foundation models, with a particular focus on healthcare, where governance, fairness, and privacy are essential.

I will share lessons learned from our team's experience developing and releasing several open-source resources, including datasets, training pipelines, and evaluation frameworks designed to support advanced AI and multimodal learning in health. These tools enable rapid prototyping, benchmarking, and reproducible experimentation across vision-language and other multimodal health tasks.

This session offers a hands-on perspective for researchers and developers looking to engage with open foundation model efforts. Whether you're releasing models, curating datasets, or contributing to tooling, you'll gain insights into best practices for documentation, data quality assurance, and aligning model development with ethical and privacy principles in health AI.

Why did we need to define Open Source AI, how did the OSI go about it, what conclusions did they reach, what's next, and how is the OSI's Open Source AI definition (OSAID) already helping to combat openwashing?

In this Session, OSI's executive director, Stefano Maffulli will present the process that led to the OSAID, and OSI EU Policy Analyst Jordan Maris will explain how the OSAID is vital to combatting openwashing in Europe and beyond.

Containers have become a cornerstone of modern IT infrastructure, yet, despite their widespread adoption, many technical fundamentals behind the container ecosystem are not widely known and lead to confusion and open questions like:

- Are containers just lightweight virtual machines?

- What are the layers of a container image?

- Are containers inherently more secure than workloads running on bare metal?

This session aims to demystify the underlying concepts behind containers and container images. We will trace the historical and technical evolution, beginning with foundational features like namespaces and cgroups, and how they are employed by container engines like Docker and Podman. Then we will provide an overview of how container images are constructed and how containers are launched from container images by Docker and Podman.

Whether you are new to containerization or looking to solidify your foundational knowledge, this talk will provide you with the necessary groundwork to successfully navigate the containerization landscape.