Arthur Mwang'onda

Open Source Summit Europe 2025 was more than a conference , it was a three-day journey through the thriving landscape of open-source technology. Held at Amsterdam’s RAI Convention Centre from August 25,27, the summit brought together developers, technologists, and community leaders from around the world collabora.com. With over 250 sessions across 15 tracks on everything from Linux and cloud to embedded systems and AI, the event offered a dizzying breadth of content milled.com. As an engineer focused on embedded and real-time systems, I mapped out a personal schedule centered on those areas (and threw in some AI for good measure). Little did I know how these themes would intertwine over the course of the week.

Day 1 , Kicking Off with Embedded Linux and Security

I arrived on Monday morning buzzing with excitement (and a dose of caffeine from the Welcome Coffee meetup). The opening talk I attended set an immediate tone: “Navigating Security Tradeoffs in Embedded Linux Systems.” Olivier Benjamin from Bootlin dived into the challenges of securing devices that have tight resource constraints. It was a pragmatic discussion about balancing robust security features with the realities of limited CPU and memory on embedded Linux devices. As someone who works with IoT gadgets, I found myself nodding along , security is critical, but so is keeping systems lightweight. This talk gave me a framework for thinking about those trade-offs (like when to use certain kernel hardening options and when a simpler approach might suffice), and it reminded me that security isn’t one-size-fits-all in the embedded world.

Directly after, I switched gears to a session on automated testing: “KernelCI for Embedded: Enhancing Reliability through Scalable Test Automation.” In this talk, Paweł Wieczorek showcased how the KernelCI project is leveraging continuous integration to test Linux kernels on many hardware platforms collabora.com. Watching real test results from dozens of boards pop up on the screen hammered home how far automation has come. For an embedded developer used to manually testing on a handful of devices, seeing scalable test labs in action was eye-opening. The takeaway was clear: investing in automated testing infrastructure (like KernelCI or lab setups) can dramatically increase the reliability of our systems , a theme that would resonate throughout the conference.

The first morning also featured the conference’s keynote sessions, and one in particular stood out: “Building Agents at Scale” by an AWS developer advocate. This keynote explored scaling AI-driven agents and tools, touching on how open source and cloud technologies enable AI growth. It was a bit of a shift from kernels and drivers, but it highlighted a major theme of the summit , AI integration. Even at an open-source event known for low-level software, AI had a presence everywhere, underscoring how machine learning and open source are increasingly linked (a point echoed by Linux Foundation leaders, who noted that open source is driving the AI boom prnewswire.com prnewswire.com ).

By mid-day, I had already rubbed shoulders with both low-level kernel experts and AI enthusiasts. Over lunch, I struck up a conversation with a developer from Linaro about supporting legacy hardware , which was perfect because the next talk I attended was “32-bit Linux Support Now and in the Future.” Kernel veteran Arnd Bergmann discussed the state of aging 32-bit architectures. It was a nostalgic and practical look at how not all devices running Linux are shiny 64-bit monsters , plenty of industrial and legacy systems still rely on 32-bit, and the community is working to keep them afloat. As someone who’s dug old ARM boards out of my drawer, I appreciated this focus. It reinforced that open source doesn’t drop support lightly; there’s a real commitment to long-term support, even as technology marches on.

The afternoon of Day 1 was a blur of networking and niche technical sessions. I popped into “Demystifying the Embedded Linux Graphics Stack”, an introductory talk on display frameworks, and later “Unpacking the Linux WiFi Stack: Writing and Integrating Wireless Drivers.” In the WiFi session, Alexis Lothoré from Bootlin peeled back the layers of the wireless subsystem. As he explained the challenges of writing drivers that play nicely with the Linux networking stack, I realized how these deep-dive topics connect back to reliability and testing. A buggy WiFi driver can crash a system , unless you have good testing and debugging in place. The emphasis on careful integration here echoed the morning’s message about quality assurance.

One unique treat was an “Ask the Expert” Q&A with Greg Kroah-Hartman, the Linux kernel maintainer. Dozens of us gathered in a lounge to pepper Greg with questions about kernel development. It’s not every day you get to ask a question directly to someone who has been maintaining Linux for decades! His candid answers (and quick wit) embodied the open-source spirit , no question was too basic, and everyone was encouraged to contribute. This informal session reminded me that behind all the code are people who truly believe in sharing knowledge. It was a fitting wrap-up for my Day 1: I walked away feeling part of a larger community, not just an attendee at a conference.

That evening, the summit hosted an attendee reception at Foodhallen Amsterdam, a trendy food market. Picture hundreds of open-source folks sampling local cuisine under one roof. I found myself at a table with developers from three different countries, exchanging stories about our favorite debugging tools over Dutch snacks. By the time I headed back to my hotel, I had a happy realization: in one day I’d learned as much from hallway conversations and new friends as I had from the formal sessions.

Day 2 , Bridging Worlds: Zephyr, Real-Time and AI at the Edge

Day 2 kicked off with fresh coffee and new topics. I was particularly excited for this morning because it delved into real-time systems. The first session I chose was “Enhancing OpenAMP: Making Linux and Zephyr Work Better Together.” As an embedded developer who dabbles in RTOSes, this talk was a highlight. Iuliana Prodan from NXP showed how OpenAMP facilitates communication between Linux and the Zephyr RTOS on multi-core systems. Essentially, you can have a Linux OS running on a powerful CPU and a Zephyr real-time OS on a microcontroller core, and OpenAMP lets them share data and messages. Watching a demo of Linux and Zephyr coordinating tasks was like seeing two worlds meet , the robust versatility of Linux paired with the predictability of a real-time system. It hit on a key theme: heterogeneous systems are becoming the norm, and open source is providing the glue (in this case, OpenAMP) to bind them. This was reinforced by the dedicated Zephyr Developer Summit track running at OSS Europe, emphasizing security, tooling, and real-world use cases for the Zephyr RTOS zephyrproject.org.

Following that, a couple of rapid-fire Lightning Talks kept the pace brisk. In ten minutes each, speakers introduced new developments in Zephyr , one on a new stepper motor driver API for Zephyr, and another on a reusable software architecture using the Zephyr message bus (Zbus). These bite-sized talks were actually great for grabbing the essence of a topic. For example, I learned how a standardized driver API will make controlling motors easier across different boards, a useful nugget for future projects. The fact that Zephyr was getting so much love at an open source Linux conference was telling: the lines between traditional Linux development and embedded RTOS development are blurring, all under the umbrella of open source.

Mid-morning, everyone gathered in the main hall for another keynote. This one, by Pallavi Priyadarshini, was titled “OpenSearch at the Linux Foundation: One Year of Open Innovation and Community Growth.” It celebrated how the OpenSearch project (an open-source search and analytics suite) had grown over the past year. While the subject was search technology, the underlying message resonated with any open source enthusiast: in just one year, a community-driven project can thrive through collaboration and transparency. It made me reflect on the numerous projects represented at this summit , each sustained by a community. Whether it’s a search engine, an operating system, or an AI toolkit, the health of the project is all about the people involved. This keynote, though not directly related to embedded systems, reinforced why we were all there together.

Back in the technical sessions, I zeroed in on safety and resilience , crucial topics for real-time and embedded folks. One standout talk was “Fail-Safe Embedded Linux: Designing for Power Resilience.” Sergio Prado walked us through designing embedded Linux devices that can withstand power losses without corrupting data. As he described techniques like using journaling filesystems and hardware watchdogs, I thought of all the times I’ve pulled the plug on a device and hoped it would reboot cleanly. Here were concrete strategies to make sure it does. The emphasis on fail-safe design tied back to the testing theme: you must test for the unexpected, be it a random power cut or a sudden surge in workload.

Before lunch, I caught a lively update in “Status of Embedded Linux”. This session, co-presented by Tim Bird and Marta Rybczynska, felt like a state-of-the-union for embedded Linux. They touched on everything from kernel trends (like real-time PREEMPT_RT developments) to community initiatives improving documentation. It was a good reality check on where things stand and where they’re headed. For instance, one insight was that despite all the shiny new hardware, some old challenges like long-term support and deterministic latency are still very much in focus. The talk gave us a sense of history and momentum , embedded Linux has come far, but there’s plenty in the works (and yes, 32-bit support is still hanging in there!).

Lunch on Day 2 was the “Better Together” lunch , a themed networking event emphasizing collaboration. I ended up at a table labeled “Embedded + AI,” which was too intriguing to pass up. Around the table, a group of us from different backgrounds brainstormed how AI and embedded might co-exist. Coincidentally, this foreshadowed an afternoon sponsor-led session I attended: “Enabling AI Agents with Real-Time Linux and Kubernetes Observability.” In that short theater talk, a tech lead demonstrated how real-time Linux patches (like the PREEMPT_RT kernel) can be used in systems orchestrated by Kubernetes, to ensure AI workloads meet timing guarantees. It sounds complex, but the demo made it tangible: think of an AI-driven robot that needs to respond immediately to sensor input , combining Kubernetes (for deploying AI models) with a real-time Linux kernel can make that possible. This was a perfect example of what I felt throughout the summit , previously separate domains (cloud, AI, real-time) are converging. And open source is the common thread enabling that convergence. I walked out of that talk buzzing with ideas about experiments to try on our lab’s robot arm back home.

One cannot talk about Day 2 without mentioning open source robotics. A session titled “An Opinionated Overview of Open-Source Robotics” by Mateusz Sadowski was both informative and fun. Mateusz gave a tour of various open-source robotics projects , from ROS (Robot Operating System) to smaller niche projects , and stressed how open collaboration is accelerating robotics innovation. For someone like me who tinkers with hobby robots on weekends, it was gratifying to see a whole talk dedicated to this field. The key takeaway was that robotics, perhaps more than any field, stands to benefit from open source: no single company can innovate in isolation on all the pieces (mechanics, control algorithms, perception, etc.), so sharing and co-developing is the smart path forward. It was a motivating message that echoed the summit’s broader theme of community.

By the end of Day 2, my brain was full but happy. I had seen Linux and Zephyr literally talking to each other, learned how to protect devices from unexpected power loss, and glimpsed how AI can be balanced on the edge with real-time constraints. The evening had some sponsored meetups and an optional city tour, but I opted for a quiet dinner with a few new friends. We exchanged cards and GitHub handles, already talking about collaborating on a small open-source side project inspired by the day’s talks. It’s amazing how quickly ideas materialize when you’re in the right environment.

Day 3 , Automation, Testing, and Closing Thoughts

Waking up on Day 3, it struck me that the summit was nearly over , yet there were still many great sessions lined up. I decided to start the final day focusing on automation and tooling, areas I’m passionate about. First on my list was “Powering Up: Lab Automation With Labgrid and CI.” This talk by Tim Orling and Trevor Gamblin showed how Labgrid, an open-source lab automation tool, can integrate with continuous integration systems to automate testing on real hardware. They demonstrated how a developer can remotely flash devices, run tests, and gather logs across a farm of boards , essentially an automated hardware lab. This was gold for me. Maintaining a test lab is often a tedious part of embedded development, and seeing how Labgrid abstracts a lot of that away was enlightening. The message was clear: automate everything you can. With tools like these, testing on physical devices starts to look almost as easy as testing in the cloud.

Next, I was drawn to a presentation with a title only an engineer could love: “Unified Boot Time Measurement: A Precise, Scalable Framework for Multi-Processor Systems.” In it, Vishnu Singh and Soumya Tripathy (from TI) introduced a framework they developed to measure boot times across complex systems with multiple processors. Now, measuring boot time might sound trivial until you consider an automotive system that has an application processor, a safety microcontroller, maybe an FPGA , all starting up in tandem. Their framework provided a synchronized timestamping method to track who comes up when, down to the millisecond. I learned some tips on instrumenting bootloaders and kernels for timing, which is something I can apply when optimizing our device boot sequences. This was a niche topic, but it underscored how far you can go with automation and measurement. If you measure it, you can improve it , and they were measuring everything about boot. It appealed to the performance geek in me.

In a shift back to hands-on practical matters, I attended “How to Support Multiple Display Controllers on One SoC.” Presented by engineers from Texas Instruments and Arm, this talk walked through adding support for a system-on-chip that had two different display output controllers. The technical challenges (different resolutions, interface standards, driver models) were discussed along with their solutions. While not everyone in the audience designs display subsystems, the broader lesson landed: embedded Linux is everywhere, even in things like digital dashboards with multiple screens, and there’s a lot of behind-the-scenes work that goes into making it all seamless. I gained a new appreciation for the graphics subsystem maintainers and also picked up some debugging tricks for GPU/driver issues.

After a quick coffee break, I found one of my favorite gems of Day 3: “Easily Generating Debian-Based Embedded Systems.” Clara Kowalsky and Felix Mößbauer from Siemens showed off a toolset for creating custom embedded Linux images based on Debian. As a long-time Yocto Project user, I was curious (and maybe a tad skeptical) about using Debian in deeply embedded scenarios. But their approach was compelling , using Debian’s robustness and package system to quickly put together an embedded OS, which could save time for certain projects. They even demonstrated building an image for a dev board in minutes. This talk reminded me that there’s always more than one way to solve a problem in open source. Yocto, Buildroot, Debian… each has its place. The key is to pick the right tool for the job and, of course, to share your approach with others. The presenters shared their recipes and scripts openly, inviting the community to try and improve them. I made a note to experiment with this when I’m back at work; who knows, it might simplify one of our project’s deployment processes.

The afternoon held one last major theme for me: integration and cooperation. A session titled “Feasibility and Architecture of a Dual-Board Embedded System: Yocto & Zephyr Integration” brought together everything I’m interested in. Andrea Ricchi and Dario Binacchi showed how they set up an embedded system using two boards , one running Linux built with Yocto, and another running Zephyr , to function as a cohesive unit. They covered communication between the boards, synchronized updates, and use cases like having a real-time controller (Zephyr) for critical tasks and a Linux system for complex processing. It was like the OpenAMP talk from Day 2 taken to the next level (with physically separate boards). This reflected a broader truth I gleaned from the summit: real-time systems and general-purpose systems are increasingly designed to work in tandem, each doing what it does best. And open source makes this collaboration possible at all levels. I was deeply impressed by how far the tooling and best practices have come for hybrid systems.

As the day drew to a close, I squeezed into one of the final lightning talks: “Protecting Zephyr Against Memory Safety Vulnerabilities With the New …” (the title cut off on my schedule printout, but the essence was clear enough!). In ten minutes, the speaker outlined upcoming features in Zephyr aimed at catching common memory bugs , think buffer overflows and null pointer dereferences , perhaps by using hardware features or safer languages. This was essentially about bringing modern memory safety to the world of microcontrollers. Considering how memory bugs are the bane of C programmers, seeing attention to this in Zephyr made me happy. It showed that even in the constrained world of RTOS, we’re adopting ideas proven in larger systems (like Rust for safety or MPUs for isolation). The fact that it was a lightning talk was almost ironic , such an important topic, distilled so briefly, but it certainly whetted my appetite to follow up on Zephyr’s security mailing list for more details.

With that, the official sessions were over. The summit wrapped up with a brief closing by the organizers thanking everyone for attending events.linuxfoundation.org. As people began to trickle out, I lingered, trying to soak in the last moments. I found myself reflecting on just how much had been packed into these three days. It wasn’t just the sheer number of talks; it was the diversity and the interconnection between topics that left a mark on me.

Before I left the venue, I wandered through the hallway one last time, where project booths were still up. I stopped by the Zephyr Project booth (couldn’t resist saying hello and snagging a Zephyr sticker) and chatted with a rep from the OpenSSF about security best practices. There was a sense of excitement and camaraderie in those closing moments , folks exchanging contact info, promising to continue discussions on Slack or mailing lists, and already looking forward to next year. In the span of a few days, we had built a little community of our own.

On the flight home, I had time to distill what I’d learned. Several key takeaways crystallized for me during this summit:

Embedded and AI are converging: No longer separate worlds, embedded systems are now running AI workloads, and open source tools are enabling this convergence. From real-time Linux powering AI at the edge to dedicated AI co-located events, it’s clear that the future of embedded includes AI and machine learning , and the open source community is leading the way in making AI accessible prnewswire.com prnewswire.com.
Testing and automation are non-negotiable: The emphasis on CI (e.g., KernelCI) and automated lab tools showed that rigorous testing is now a cornerstone of embedded development. We saw how projects like KernelCI and Labgrid can catch issues early and improve reliability collabora.com. I’m convinced that investing time in automation will pay off in fewer headaches down the line.
Security remains front and center: Whether it was securing devices under resource constraints or planning for new regulations, security was a recurring theme. Open source projects are proactively tackling security , from memory-safe frameworks in Zephyr to foundational secret management tools (as I noted at the OpenBao booth, where demand for open security solutions is high openbao.org). The community is treating security as a continuous journey, not a one-time fix.
Open source is about people and community: The most profound lesson wasn’t technical. It was seeing the human side of open source , maintainers mentoring newcomers, companies collaborating on neutral ground, and the genuine enthusiasm everyone had to share knowledge. One quote I saw from the event encapsulated it well: “Open Source Summit is a fundamental gathering place for exchanging ideas across projects and meeting all of the people who make open source communities work.” events.linuxfoundation.org I felt that truth in every hallway chat and during every Q&A. It’s the people that make the code possible.

In the end, Open Source Summit Europe 2025 felt like a celebration of collaboration. I came away not only with new techniques to try and problems to solve, but also with a reinforced conviction in the open source approach. The summit showed me that whether we’re dealing with a tiny microcontroller or a cloud AI cluster, we move faster and build better when we do it together openly. As I return to my daily work, I’m carrying that spirit with me , along with a notebook full of ideas and the contacts of some amazing engineers I can now call friends.

Embedded systems, real-time Linux, AI integration, security, automation , all these pieces are coming together to shape the future of tech, and open source is the glue. Attending OSS Europe 2025 was a thrilling glimpse into that future. I’m already looking forward to the next gathering, but until then, it’s time to hack on some of these new ideas. If there’s one thing I’m sure of, it’s that the conversations started in Amsterdam will continue on mailing lists, code repos, and forums, driving innovation long after we’ve left the conference halls. And personally, I can’t wait to be part of that ongoing story.