BITS&CHIPS EVENT 2025

Software is everywhere, and (almost) everything we do relies on software. But can we actually rely on software? Software bugs are just as old as software and frequently cause major disruptions, such as the recent Crowdstrike’s outage due to a software update.
I’ll argue that it should be possible to improve this situation by using formal methods. Formal methods can help identify software bugs in the early stages of development, so that they can be identified and fixed before deployment. They can also be used to analyze the possible effects of software updates before applying them.
Of course, the use of formal methods requires a time investment and training. However, the return on investment is big because of the improvements in software quality. I’ll end by discussing what needs to be done to make the transfer of formal methods from academia to industrial practice.

Unlock the secrets to building truly secure and resilient products. The escalating value of embedded algorithms and designs is making manufacturers prime targets for hardware IP theft by competitors and hostile state actors. Simultaneously, new regulations like the Cyber Resilience Act (CRA) are creating a mandatory requirement for security to be integrated into product design.
The core strategy is a “shift left” approach, embedding security at the start of the development cycle. Since products are deployed in potentially hostile environments, security must be rooted in trusted, tamper-resistant hardware, not just software. Learn how formal verification, side-channel attack prevention and fault injection resistance can be practically applied. Discover how embracing design-level security is a strategic investment that preserves IP, ensures critical market access and dramatically reduces costly lifecycle management.

Several analyses of successful and unsuccessful development projects highlighted two key success factors for us: timely access to experts who drive both high innovation and early risk reduction, and the critical role of technical leadership. Focusing on the latter, we assert that first- and second-level engineering leaders are pivotal in shaping an environment conducive to high engineering performance. Beyond the orchestration of technical problem-solving, these leaders must grasp business models, cultivate the engineering culture, build diverse teams, integrate new technologies and translate cost and time pressures into innovative working methods, driving continuous transformation.
To cultivate these essential leadership skills, we developed the Leadership Focus Group, a guided workshop format. The core involves participants crafting individual leadership models, fostering transparent discussions on goals and expectations. Through individual and group coaching, leaders explore and verbalize their convictions, enhancing self-awareness and understanding of colleagues. These intensive workshops are framed by theoretical training on “Leading of X,” where X can be systems engineering, artificial intelligence or any further engineering technology, work organization or change model.
This challenging yet rewarding process of formulating a coherent leadership vision, reviewed with peers, consistently leads to significant personal and professional growth. The overwhelming positive feedback has led to global adoption, reaching over 7,000 participants and establishing a robust foundation for future challenges.

At Fontys UAS, we’ve been educating software engineers for many years. Since 2019, we also educate AI engineers: software engineers who know how to build production-ready AI systems. Recently, we saw a whole new combination of AI and software engineering: using LLM-based tooling to automate the software engineering process itself. We quickly adapted by adding new semesters to our curriculum, starting employee professionalization programs, building necessary infrastructure, but most of all by starting experiments. We educate the AI-augmented software engineer of the future. Working toward this moving target has given us valuable insights into how to transform your software engineering processes in practice and how the job of a software engineer might change. We’ll share our lessons learned both for individual software engineers and architects worrying about their jobs, as well as for engineering companies worrying about their future.

How do you embed AI into the DNA of a global manufacturing organization? At Canon Production Printing, we believe it starts with empowering every employee to explore and initiate AI in their daily work. As chair of the Center of Expertise for AI, I’ll share how we’re scaling digital transformation across R&D, manufacturing and support functions. Our journey began with inspiration: AI Days that sparked curiosity and lowered barriers. Then came structure: a cross-sector Center of Expertise that facilitates exploration, supports project initiation and fosters shared learning. I’ll walk you through our approach, the cultural shifts we’re nurturing and the lessons we’ve learned in making AI not just a tool, but a mindset. I’ll also share our experience applying AI in software engineering, from accelerating development to enhancing quality. Whether you’re just starting or scaling your AI journey, this talk offers practical insights into building momentum and making transformation stick.

Formal methods may be utilized in software development processes to guarantee the absence of coding guideline violations and runtime errors, and to provide safe bounds for worst-case stack usage and worst-case execution time. For highly safety-critical applications, using ‘sound tools’ that implement such formal methods is well established. The experiences gained over the past years show that a formal approach not only greatly increases confidence in the correctness of code but also significantly reduces qualification efforts and costs in development, production and maintenance.
The benefits of sound tools are, however, not limited to highly safety-critical applications only. At Rolls-Royce Solutions GmbH (formerly MTU Friedrichshafen), their well-established sound toolchain (introduced for their first generation of electronic engine controllers for emergency diesel generators in civil nuclear power plants) is now being extended to be applied to all embedded software development projects.
This talk presents the lessons learnt at Rolls-Royce Solutions GmbH when migrating their software development processes to using sound analysis tools. It demonstrates with concrete example projects how a sound toolchain enables supporting novel requirements imposed by the Cyber Resilience Act (CRA). It also discusses how using sound tools complements basic rules of the ISTQB.

The manufacturing sector ranks among the top three targets for cyberattacks. While digitalization boosts business opportunities, it also introduces cybersecurity challenges. Resolving these challenges is a prerequisite for a successful transition to smart factories. Yet, manufacturers struggle to find the right balance between cyber resilience investment and avoiding overspending. Furthermore, no one-size-fits-all solution seems feasible, since the digitalization maturity across manufacturers varies widely and progresses over time. From initial administrative digitalization to digital twin exploration and ultimately to lights-out operations.
To face this challenge, we present the Cyber Resilience Blueprint Methodology (CRBM). CRBM guides manufacturers in selecting appropriate security measures based on their business risk appetite, their progress in transitioning to smart factories and their threat profile. In this talk, practical cases are used to illustrate how CRBM serves as a stepping stone toward autonomous cyber-resilient factories.

In today’s complex products, software is a core part of the system itself, and its behaviour, timing and evolution shape the broader system just as much as physical components do. As the head of a software department in a ‘hardware’ company, I’ve found that many of the most critical conversations I have with software designers and architects aren’t about code, patterns or APIs, but about system context: purpose, constraints, integration and trade-offs.
This talk explores why systems thinking is essential for software minds and how cultural heritage subtly shapes the way we approach it. By stepping beyond the code, software architects gain clarity on the “why” behind the “how” – leading to more resilient, adaptable and meaningful designs. Likewise, system architects must understand the nature of modern software systems to create coherent, sustainable solutions. Bridging the gap isn’t optional anymore; it’s the key to building the next generation of complex, reliable systems.

AI is redefining software engineering, fast. This leaves us with urgent questions: Is our profession becoming obsolete? Can anyone build software with the help of AI? And what separates hype from reality? Drawing on practical experience from over fifty software engineers at Kabisa, this talk helps you cut through the noise. You will leave with a realistic view of AI’s current capabilities and limitations in our field, the latest tools and techniques to effectively integrate AI into your development processes and actionable advice for how the software engineer role must evolve to stay not just relevant, but become even more impactful than it is today.

Discover how Philips IGT-Systems is pioneering the use of AI to accelerate innovation and boost productivity in R&D. In this talk, we’ll explore how emerging technologies like large language models (LLMs), retrieval-augmented generation (RAG), Model Context Protocol (MCP) and AI agents are transforming the way we work – streamlining documentation, enabling faster insights and supporting engineers in their daily tasks. You’ll get a practical look at the early applications we’ve implemented, lessons learned and how AI is helping us shift from manual, repetitive work toward high-value innovation. Whether you’re new to AI or already experimenting, this talk will provide a clear, engaging primer on the core building blocks – and a behind-the-scenes glimpse into how we’re embedding them into a complex, highly regulated R&D environment.

In this talk, we’ll explore software testing through the lens of probabilistic sampling. Our main focus will be on finding concurrency bugs, which arise from the unpredictable behavior of concurrent and distributed systems. Detecting these bugs is notoriously challenging because they often occur only in specific, subtle interleavings of the events. While naive random testing is often ineffective at uncovering challenging bugs, our randomized testing algorithms offer effective methods by strategically sampling test executions.

The past five decades have seen an exponential increase in compute chip performance. In the meantime, progress in memory performance has usually lagged behind. Historical memory technologies such as SRAM, DRAM and 3D-NAND have forced the appearance of a memory hierarchy to deal with the inherent trade-offs between density, speed, endurance and energy consumption of these technologies. The past three years have seen a surge of generative AI workloads, which have brought the memory requirements far beyond what the current memory hierarchy can support. The results: extremely expensive and complex high-bandwidth memory (HBM) chips, combined in extremely expensive and complex AI compute systems. This happens at a time when the fundamental scaling roadmaps of most of these technologies are screeching to a halt.
In this talk, we’ll review the state of generative AI and the hardware context. We’ll then present Vertical Compute’s technology, which exploits the 3rd (vertical) dimension to integrate bit strings vertically, directly above transistors. Direct CMOS integration in the transistor back-end means no more chip-to-chip data shuttling: no bandwidth bottleneck and much lower system power. Combined with high-performance transistors, such architecture will enable a completely new class of products capable of processing generative AI workloads on the client side. The benefits? Private and secure (personal) generative AI inference at the user’s fingertips. We expect hardware leaps, such as the one proposed by Vertical Compute, will be the way to truly democratize generative AI applications and unlock their full potential.

Systems architecting is the process of conceptualizing, designing and leading the development and engineering of complex technical systems. For whom? Who are the main stakeholders? The landscape of opinions, requests and needs can be a minefield, with input coming from every possible direction and beyond! This includes a mixture of internal and external opinions – multiple dozens of them – none of which are aligned. It seems like an impossible endeavour. And, of course, we mustn’t forget your own most important opinion: the need to articulate that brilliant idea, that beautiful piece of technology, that symphony of flow, energy and movement, that incredible piece of art that you’ve shaped and modeled in your mind.
High-tech veteran Ger Schoeber will go head-to-head with the audience, playing devil’s advocate to established systems architecting views – and he’s inviting you to join the debate.

In our data-driven world, from manufacturing floors to financial markets, identifying anomalous or abnormal behavior is crucial. But how do you detect critical equipment malfunctions or fraud when sensor data is noisy, high-dimensional and stochastic in nature? Traditional methods often fall short. Join our talk to see how we fuse the domains of digital signal processing and deep learning to accurately and efficiently tackle this challenge.

How do you introduce generative AI in an environment where data can’t move, cloud isn’t an option and everything is subject to strict regulations? In this talk, Pepijn Noltes shares his experience setting up local generative AI solutions within a defense context. Through two concrete pilots – from GPU-powered developer machines to a Mac Studio cluster – he demonstrates how innovation is still possible within closed systems. He also discusses the role of the Local GenAI Board, which uses the RICE method to prioritize use cases. A talk full of practical lessons, technical decisions and organizational insights for anyone looking to deploy AI in a restricted environment.

The world of software testing is undergoing a radical transformation. And AI is leading the charge. Once ruled by scripts and repetition, quality assurance is evolving into an intelligent, adaptive discipline. In this talk, we explore how AI isn’t just accelerating testing, but redefining what it means to test altogether.
From predictive defect detection to self-healing automation and AI-generated test cases, we look at real-world applications that are accelerating delivery and improving quality at scale. But it’s not just about tools; it’s a mindset shift. What does this mean for test engineers, QA leads and the broader development lifecycle?
What skills will tomorrow’s testers need to thrive in an AI-augmented world? Join us on this thought-provoking journey and discover how you can ride the AI wave instead of being swept away by it. The future of testing isn’t coming; it’s already here.

Billions of sensors power our smart devices, a number that’s set to explode in the years ahead. More sensors mean more data, and more data demands smarter, leaner AI. Neuromorphic computing has an answer to this challenge: spiking neural networks (SNN). Inspired by the brain’s own mechanisms, SNNs stand out with their sparse activity, event-driven processing and natural handling of time. They’re the brain’s playbook for energy-efficient intelligence. But here’s the twist – the magic isn’t just in the networks; it’s in the system.
In this talk, we’ll dive into neuromorphic computing at the sensor edge and introduce Innatera’s Pulsar, the world’s first neuromorphic microcontroller. Pulsar brings brain-like AI right to sensors, unlocking massive gains in battery life, performance and device functionality. With real-world applications, we’ll show how neuromorphic tech is reshaping the future of smart devices – and where it’s headed next.

Imagine a development team always on standby, no salaries, no sleep, just power and cooling. Planning, coding, testing and documentation, all triggered by simple instructions at your keyboard. This is possible when we let language models collaborate as agents.
Large models like ChatGPT have shown they can reason and complete complex tasks. Using those proprietary models means giving up control over your data and IP. That’s no longer a trade-off you have to make. With small language models (SLMs) running locally and working together as specialized agents in a multi-agent system (MAS), you can build autonomous workflows securely, without compromising performance.
In this talk, we’ll show how multi-agent systems with small language models can be used to automate complex tasks while maintaining control over data and access. We’ll walk through real-world scenarios and practical tools, so you can stay ahead before multi-agent systems become the new standard.

Imagine a world where your system doesn’t wait for alerts – it calls you first. Instead of reacting to failures, agentic AI anticipates them. Weak signals trigger smart actions before downtime strikes. This talk reveals how Vanderlande Digital Services is redefining proactive service: detecting anomalies early, automating root-cause analysis, coordinating responses across teams and scaling expert guidance with AI agents. Discover how we’re transforming service models, boosting uptime and building trust – while reshaping the role of service engineers. Curious about a future where systems think ahead? Join us and see what happens when your system calls you first.

What happens when systems engineering meets national character? From US-designed deepwater floaters offshore Borneo, to floating LNG co-designed with Japanese engineers, and from Italian oilfield design to Dutch and English collaboration – this talk explores how cultural norms shape engineering decisions, communication and hierarchy.
Expect sharp contrasts: British managers who master understatement but fear upward delegation, Japanese teams where initiative doesn’t exist and American engineers who comfortably silo themselves. Italian meetings? Numbers first, boss speaks, everyone nods. And the Dutch? Informal chats, technical heroes and decisions made at the lowest level – often before anyone realizes.
Drawing on real-world experience from Shell and global projects, this talk offers a humorous yet insightful look at how culture influences systems engineering and architecture. Whether you’re engineering systems across borders or just trying to decode your team’s behavior, this talk should leave you thinking differently about collaboration.

To meet the goal of net-zero greenhouse gas emissions by 2050, the Netherlands must scale its electricity grid by a factor of five. This talk explores how product line engineering, a well-established approach in high-tech, can accelerate grid expansion and thus the energy transition. We’ll present the sector’s modular building practices and the engineering challenges involved in expanding the electricity grid. We’ll compare these with high-tech product development approaches and highlight transferable engineering principles as well as sector-specific constraints. Attendees will gain insights into how system approaches from the high-tech sector can enhance strategies for addressing one of today’s major infrastructure challenges.

Generative AI is redefining how we build and manage software. What began as simple code completion has evolved into a fundamental shift in how development teams design, test and deliver solutions. This talk explores the rapidly advancing landscape of Gen AI for software engineering, highlighting emerging capabilities, benchmark insights and the changing role of developers in an AI-assisted environment.
We’ll discuss how applying Gen AI design patterns from retrieval and reflection to collaboration and governance can unlock new levels of productivity, quality and adaptability across the software lifecycle. The focus will be on understanding where these models excel today, where they fall short and how to integrate them responsibly and effectively into modern engineering workflows.
Join us to discover how the next generation of software development is being shaped by Gen AI where human creativity and machine intelligence come together to transform the way we design, build and evolve technology.

In our journey to integrate AI into software development, we’ve encountered a landscape that ranges from quick wins to complex challenges. This talk explores that spectrum. We started with low-friction successes like deploying AI coding assistants (eg Github Copilot). We then looked at other low-hanging fruits like querying large semi-structured data and fixing simple code smells. As we progressed, we faced more nuanced technical hurdles: deeper code refactoring, collecting high-quality validation data and the limitations of current models in understanding context-rich codebases. The real complexity, however, emerged not from the technology but from the organization. Inconsistent adoption among engineers and the contextual mismatch between AI tools, real-world workflows and addressing IP security and compliance concerns all posed significant barriers. Not to mention, there needs to be proof of return on investment and due consideration for sustainability and environmental impact. This talk distills our experience into key takeaways that reflect both the promise and the responsibility required to make AI truly impactful in software engineering.