Klaus Skafte
DenmarkKlaus Skafte works at the Danish Agency for Digital Government where he ensures that the suppliers used for the big Digital Infrastructure projects deliver on the quality needed. It sounds easy – the supplier just need to do as the contract requires of them, but in reality it is not an easy task.
In addition, Klaus is part of exec for DSTB (the Danish member board of ISTQB), when he is not busy building Lego, reading comic books, working on home automation… the list goes on and on.
Did anybody hear that?
"In space no one can hear your scream", this is a quote from a movie and it may hold true within operations too.
Often alarms meant to monitor applications that are vital to a company, or even country, are not configured correctly - but why didn't we test them?
Grzegorz Holak & Filip Barszcz
PTB/Sii Poland, PolandGrzegorz is QA Architect and AI Ambassador with over 10 years of industry experience, specializing in test automation, mainly in fintech projects. A speaker with dozens of presentations at conferences, meetups, and industry events in Poland and abroad, where he shares knowledge in the areas of software quality and testing. Actively involved in the QA community as a member of programme committees, co-creating agendas and selecting topics relevant to the development of the industry. Supports testing teams in implementing GenAI solutions and modern tools by delivering talks, workshops, and educational materials. Combines a solid approach to test automation with experimentation at the intersection of QA and AI, focusing on solutions that genuinely accelerate the test creation process. After hours, he develops his own projects: teslawbloku.eu (PL only), quality-blog.eu and ai-slop.win.
Santander Corporate and Investment Banking, Poland
Throughout his career, Filip has collaborated with renowned organisations such as SCIB (Santander Corporate & Investment Banking), T-Mobile, Capital.Com, and IQVIA.
He specialises in building and refining quality assurance processes, mentoring QA professionals, and fostering close collaboration between QA and development teams.
As a strategic QA leader, he has driven major organisational transformations — from building QA departments from the ground up to restructuring teams for greater efficiency and alignment with business goals.
He has successfully defined test strategy, designed automation architecture, and implemented multi-level testing - from unit to end-to-end coverage.
Always curious and forward-looking, he continuously explores emerging tools such as AI-driven test automation and predictive defect analysis.
Outside of work, he is a passionate storyteller and avid enthusiast of TTRPGs, LARPs, miniature painting, and science fiction literature.
AI-Powered QA: From Prompt Frameworks to Multi-Agent Workflows
Most QA professionals already use AI tools like ChatGPT or Claude in their daily work - but without structure, the results are inconsistent, shallow, and hard to reproduce. This 6-hour hands-on workshop bridges the gap between casual AI usage and professional prompt engineering tailored specifically for software testing.
Participants will learn and immediately practice 10+ prompt frameworks (CRISPE, Chain of Thought, Tree of Thought, RAG, Few-Shot, CO-STAR, SCQA, and more), each mapped to concrete QA tasks: generating test cases, analyzing edge cases, writing bug reports, building business cases for automation, and creating test code from API documentation.
The workshop goes beyond single-model prompting. Attendees will explore multimodal capabilities - using image, audio, and video inputs for bug reporting and UI analysis - and learn how Model Context Protocol (MCP) enables AI agents to orchestrate real tools like Jira, Confluence, GitHub, and Slack in unified workflows. We’ll close with multi-agent scenarios where different AI tools collaborate on complex QA tasks: from incident response to framework migration.
Every concept is reinforced with live exercises. Participants leave with a personal prompt library, ready-to-use templates, and a clear understanding of which AI tool to use for which QA task.
Valery Penev
Adastra, BulgariaValery Penev brings over a decade and a half of experience in data warehouse consulting and data engineering at Adastra. Throughout his career, he has worked across diverse industries, clients, and roles. In recent years, his focus has expanded to include cloud technologies. He has also recently started his PhD journey, further deepening his academic and professional development. Valery is goal-oriented and analytical, with excellent interpersonal skills. For eight years, he served as a Talent Manager, mentoring a team of ten, and since 2016 has lectured at the Adastra Academy. He spent almost a decade playing a key role in technical interviews as part of Adastra's hiring process
In 2020, he co-founded "Out of the Box Ltd." - a company focused on web services, digital marketing, and testing services, helping small and mid-sized businesses in building a strong online presence.
His superpower is to be creative and always positive.
The Future of Testing Starts With You
As AI, automation, and machine learning accelerate, it's easy to think that testing is losing its relevance - that tools, bots, and algorithms will soon take over our craft. But the truth is, the real future of testing isn't about technology - it's about people.
This talk explores why the testers of tomorrow will need more than just technical expertise. Curiosity, creativity, empathy, and critical thinking will become the superpowers that separate great testers from good ones. Through practical examples and stories from modern testing teams, we'll look at how human qualities drive better collaboration, smarter automation, and deeper product understanding.
It's a call to action to grow, adapt, and lead with a mindset that embraces both innovation and humanity. Because while tools may change, the essence of testing - asking questions, seeking truth, and caring about quality - remains deeply human.
The future doesn't just happen to us - it starts with us.
Gjore Zaharchev
Avenga, North MacedoniaGjore Zaharchev boasts over 19 years of expertise as an Agile Evangelist and a dedicated Heuristic Testing advocate. His extensive background encompasses Automated, Manual, and Performance Software Testing across diverse domains and clientele. Throughout his career, Gjore has demonstrated leadership by overseeing QA personnel and teams dispersed across Europe and the USA, managing varying team sizes. Recognizing testers as individuals equipped with diverse problem-solving skills and an engineering mindset, he firmly believes that Software Testers go beyond being mere statistics for clients. Currently serving as the Head of QA for SEE region for the Software Testing Team at Avenga, Gjore is committed to ensuring top-notch quality. In addition to his managerial role, he holds the position of an esteemed speaker at numerous conferences and events across Europe, and he serves as a Testing Coach at Avenga Academy in Skopje. Notably, since 2020, Gjore has contributed as a board member for Macedonia at SEETB, an ISTQB-affiliated organization.
Since 2020 he is SEETB (an ISTQB) board member for Macedonia.
AI-Assisted Testing in Practice: Separating Real Productivity Gains from Automation Illusions
Artificial Intelligence has rapidly entered the software testing toolbox. Tools now promise self-writing tests, self-healing automation, and even autonomous QA pipelines. Yet teams experimenting with AI-driven testing often encounter a different reality: increased flakiness, unreadable automation, and test suites that quietly accumulate technical debt.
This session presents a field-tested perspective on AI adoption in modern Quality Engineering, based on real implementation patterns observed in production environments.
Through a series of real-world examples and practical workflows, the session explores how AI can significantly accelerate specific testing activities, including:
However, the talk also examines systematic failure patterns observed in AI-driven QA initiatives, including:
Instead of asking whether AI will replace testers, this session addresses a more important question:
How can testing professionals use AI as a force multiplier for disciplined engineering teams?
Attendees will leave with concrete techniques and decision frameworks that allow them to leverage AI responsibly while preserving the reliability and intent of their testing strategy.
Wim Decoutere
CTG, BelgiumWim Decoutere is a master in informatics who started his testing career at CTG Belgium almost 15 years ago and has been testing at a number of projects ever since, mostly in the financial sector. Wim feels at home when standing in front of a classroom. Since he became a full-time trainer, he has taught hundreds of people about the wonderful worlds of testing and requirements engineering. As a veteran youth instructor with a passion for learning theories and people management, Wim is constantly looking for new ideas to improve his own performance and that of the entire testing team.
Wim is secretary of the BNTQB and an associate member of IREB.
Exploratory Learning Styles
Cem Kamer defined Exploratory Testing as ‘simultaneous learning, test design and execution’.
If you would ask a group of testers to define Exploratory Testing, most of them do fine on the simultaneous test design and execution, but they struggle with the simultaneous learning. This does not come as a surprise to me. If you would ask that same group of testers to define the concept of learning, the result is usually incoherent murmuring.
Even though everybody knows what learning is, it is hard to define or explain it. You can compare it to trying to explain how to walk. Luckily, there are some brilliant people on this planet, like David Kolb et al., that managed to define ‘learning’. They even identified different styles of learning. Every person has a preferred learning style, just like every tester has their own set of heuristics to test an application. Knowing your preferred way of learning an application helps in coming up with even better tests while exploring. Being aware of other learning styles will even expand your exploratory learning abilities.
In his talk, Wim takes you on an exploratory journey to discover your own preferred way of learning. Based on the recommendations from well-established learning theories, he’d like to point out how you can apply these learning styles to get more out of your exploratory testing and to learn more from your test subjects to improve your testing.
Victor Ionascu
Axway, RomaniaVictor has 15+ years of extensive experience in experimenting, learning from failures, and helping others think outside the box. Currently, he works on integrating multiple products into high-quality solutions at Axway. He has spoken at many international conferences, sharing his love for eliminating unnecessary tasks and focusing on what truly adds value. Outside of work, he enjoys hiking, motorbiking, climbing mountains, and spending time with his three kids.
assertEquals() Is Not Enough
Artificial intelligence is no longer a separate system. It is now embedded inside enterprise software — generating mappings, assisting transformations, writing rules, making decisions. But most teams are still validating these AI-assisted features using deterministic testing strategies designed for traditional systems.
This talk explores what breaks when assertEquals() is no longer a reliable oracle.
Using a concrete enterprise scenario — a financial integration platform with an AI-powered Mapping Copilot that generates DML → XML transformations — the session walks through real validation gaps that traditional QA would miss. The generated mappings pass XSD validation. The pipeline is green. Yet subtle semantic errors can silently corrupt financial data.
The session examines why regression assumes reproducibility, stable outputs, and fixed expectations — and why AI systems violate all three. From hallucination to drift to adversarial prompts, the failure modes are different. Therefore, the testing model must change.
Attendees will learn:
This talk is valuable for QA engineers, automation engineers, architects, and team leads working with AI-enhanced systems in enterprise environments. It provides concrete, implementable patterns — not theory — and bridges traditional QA with emerging AI reliability practices.
Tal Pe'er
Grove Software Testing, NorwayTal has been working in software and system testing for over 25 years, getting his Foundation Certification back in 1999. He’s been working as a tester and test manager before becoming a trainer and consultant, helping organizations to establish and improve test teams and test processes, as well as training testing professionals with various ISTQB® courses and testing workshops.
Tal has been active with ISTQB® since 2008 and has been a member of the Executive Committee between 2017-2023.
Tal is a Principal Consultant at Grove Software Testing, one of UK’s leading training providers and course training materials.
AI Generates Tests, Humans Generate Confidence
IAI is changing testing faster than any tooling shift we have seen in years. Today, AI can generate test ideas, build automation infrastructure, summarize failures, and support rapid exploratory sessions. Teams are also experimenting with lightweight techniques such as vibe testing, using AI prompts and intuition-driven tests to quickly validate whether a feature appears to work.
This creates speed, but it also creates a dangerous illusion. The easier it becomes to generate tests, the easier it becomes to generate evidence without depth.
In this session, I will draw on my experience from modern Agile teams to show how the tester’s role is shifting from test creation to confidence engineering.
I will show how even with AI-generated tests, green pipelines, and even vibe testing all passed, critical production defects still escaped.
This is not an anti-AI talk. It explores why human testers remain essential. Not because humans are better at generating volume, but because humans excel at:
• questioning assumptions
• recognizing weak oracles
• understanding business risk
• exploring edge conditions
• spotting when green does not mean safe
The session introduces a practical confidence framework attendees can apply immediately in their own teams:
• use AI to generate evidence
• use humans to challenge assumptions
• use risk to decide sufficiency
• separate “working” from “trustworthy”
Petko Petkov
Dreamix, BulgariaPetko Petkov is a senior quality assurance engineer with more than 7 years of experience, interested in automation testing, automation of tasks, improving the efficiency of the teams even when it was not requested by the client, and supporting others to develop as professionals (1:1 mentorship, preparing workshops, writing articles and having presentations). Joining Dreamix in 2020 he had the chance to take part in several different projects where he helped with the CI/CD implementations, automation of test creation and execution within test management systems, and preparing custom solutions for testing NLP software. In addition to that Petko’s interests in philosophy led him to becoming a PHD student in contemporary philosophy in the Sofia University.
Automating Asynchronous Events
Do your automation tests fall apart when facing unpredictable timing events? Is your team still manually testing critical asynchronous functionality? Discover how to conquer one of QA's most challenging frontiers - asynchronous events - and revolutionize your testing approach!
In the world of test automation, we've mastered many challenges - but asynchronous events remain the final frontier that separate good automation frameworks from great ones. When cron jobs run on their own schedule, emails arrive "eventually," and file generation happens behind the scenes, traditional automation approaches fall short.
Many teams reluctantly resort to manual testing for these scenarios, introducing inconsistency and delays that undermine the benefits of your automation investment. But what if there was a better way?
By the end, you won't just understand asynchronous automation - you'll have expanded your professional toolkit with versatile solutions that can be applied across projects, technologies, and industries. These are the skills that separate average QA engineers from the automation experts who drive innovation and efficiency.
Istvan Forgacs
4-Test Plus, HungaryIstván Forgács, PhD, began his career as a researcher at the Computer and Automation Research Institute of the Hungarian Academy of Sciences. He has published numerous scientific articles in leading international journals and conference proceedings. He is the lead author of the books Modern Software Testing Techniques (Apress), and Practical Test Design (BCS), and the co-author of Agile Testing Foundations (BCS).
In 1998, István transitioned from academia to entrepreneurship by founding Y2KO, a startup that provided an efficient solution to the Y2K problem.
He is also the founder and CEO of 4Test-Plus. István is a notable figure in the software testing community. He has served as an author for the Advanced Test Analyst Working Group and was a member of the Agile Working Group at ISTQB. He developed an intelligent mutation testing framework to help testers enhance their test design skills. Additionally, he is the creator and key contributor of Harmony, the only two-phase model-based test automation tool. He was a keynote, invited, and regular speaker at several academic and industrial conferences.
Automate Tests Better, Faster, Cheaper
Most test automation tools focus solely on automating test execution, often overlooking the critical aspect of test design, which means they can only detect simple bugs. Model-based testing (MBT) addresses this gap by incorporating test design into the process. However, current MBT methods face several challenges:
• Coding Requirements: They necessitate guard conditions, meaning testers must engage in coding.
• Complex Output Modeling: Modeling outputs remains a difficult task.
• Learning Curves: Testers need to adapt to a model editor, which can be time-consuming.
A new model-based technique, Action-State Testing, effectively overcomes these challenges. This specialized modeling approach comprises a potential initial state, labels, and action-state (model) steps. Labels encapsulate requirements and model elements, while model steps define the test steps. By utilizing states, models become more concise and capable of detecting most defects.
In this method, model steps are represented as nodes within a graph, described through simple text. Each step begins with an action (input), followed by zero to multiple responses (output), with the option of specifying an arriving state. For example:
add items for 40 euros ⇒ total price is 40 ⇒ free beverage can be selected STATE free beverage offered
Steps can be sequentially connected within the same test if there is an edge from one to another. A fork occurs when there’s an edge from step 0 to both step A and step B, resulting in two separate test cases (step 0, step A) and (step 0, step B). After forking, common steps may require execution for both test cases, effectively joining the forked paths. Using a text editor simplifies the addition of fork and join model steps.
Advantages of Action-State Testing:
• Entirely codeless
• Inherently adheres to the DRY (Don't Repeat Yourself) principle
• Quick model creation
• Simple and cost-effective maintenance
• High Defect Detection Percentage (DDP)
Action-State Testing has been rigorously evaluated using a specialized intelligent mutation framework, which you can explore at https://test-design.org/practical-exercises/. By comparing action-state methods with AI, manual testing, and stateless solutions, we observed that their DDP ranged from 40-80%, whereas action-state testing achieved a perfect 100% DDP, effectively killing all 65 mutants. This method has already been applied to numerous real-world projects, identifying several defects. The average time for test design and execution automation of a single test case is less than one hour—a significant improvement over the industry average of over four hours without systematic test design.
In my presentation, I will demonstrate the functionality of the action-state testing tool and how to implement test cases from the model. I will also address key challenges we've encountered, such as flaky tests, dynamic selectors, and code changes. Additionally, I will illustrate why action-state testing is particularly effective for identifying intricate bugs using the new test version of the ISTQB Glossary application.
Dmitriy Sobko
EPAM,RomaniaI am a Quality Architect with 14+ years of experience delivering end-to-end quality solutions for complex enterprise systems. Throughout my career, I have combined hands-on expertise in test automation and quality engineering with leadership roles focused on enabling teams to build scalable, resilient, and maintainable software ecosystems. My areas of interest include AI-driven testing, Big Data quality assurance, cloud-native architectures, and the evolving role of quality in modern software delivery. I enjoy exchanging ideas and sharing lessons learned from real-world projects with peers at industry events and within professional communities.
AI Agents Driven Testing: Redefining Quality at Scale
Testing is entering its next era: from automation scripts to autonomous quality agents. While many teams use AI as a coding assistant, few leverage its full potential to participate actively in the testing lifecycle.
In this talk, you'll discover how AI agents equipped with reasoning, memory, and tool access can support real-world QA activities. Through examples powered by AI/Run CodeMie, we'll explore the design and governance of agentic testing systems and showcase three practical agents: a Test Case Creation Agent, a Test Data Generation Agent, and an Automation Test Creator Agent.
Join this session to learn how to move beyond copilots and build the foundations of an agent-driven quality ecosystem.
Ahmed Hassan
NTT Data, GermanyAhmed Hassan is a Lead Test Architect and AI Engineer at NTT DATA with 15 years of experience in software testing, test automation, quality engineering, system analysis, and project coordination. He is based in Munich, Germany, and holds a Bachelor's and Master's degree in Computer Science. Ahmed has worked across several industries, including automotive, manufacturing, financial services, telecommunications, logistics, and utilities. He is also a speaker at several international conferences across Europe and Africa.
Architecting Intelligent Trust, Building a Digital Immune System for Modern Testing
In today's world of AI code generation, microservices, and CI/CD pipelines, software delivery happens at speeds too fast for any traditional testing strategy. In this reality, running full regression testing suites is not a way to assure quality anymore - rather, it causes delays, flaky tests, and false sense of security.
In this presentation, we introduce you to the concept of Quality Operating System, a modern architecture to transform testing into an intelligent trust layer throughout the entire software lifecycle. It operates differently than your classic test runner by analyzing code changes, historical defects, and runtime data to execute just those test paths that matter.
We'll guide you through an actual scenario, when each new commit launches a chain reaction of risk evaluation, focused test execution, and ongoing observations of the application to optimize further testing. We'll also reveal how our concept combines your favorite testing tools like Playwright, CI/CD pipelines, observability platform with a powerful orchestration engine connecting them all together.
We believe this architectural vision will inspire you to move from a reactive model of testing into an intelligent self-optimizing process that reduces testing execution time, decreases amount of flaky tests, and increases confidence in releases.
Anastasia Engelhardt
Luxoft, SerbiaAnastasia Engelhardt is a Quality Assurance Engineer, public speaker, Women Techmakers Ambassador, and tech blogger focused on exploratory testing, QA mindset, and practical software quality. Beyond her engineering work, Anastasia contributes to the Balkan tech community through mentoring, educational content, workshops, public speaking, and women-in-tech initiatives. Through her blog, she shares practical insights about Quality Assurance, career growth in IT, and the evolving role of testers in the AI era.
How to Build a Manual Evaluation Framework for AI RAG Chatbots Without Code Access
In this presentation, Anastasia will share a practical QA approach to evaluating AI-powered RAG chatbots without direct access to the code, model, or internal retrieval logic. The talk focuses on how testers can build a manual evaluation framework, define answer quality criteria, assess groundedness and context handling, identify release risks, and communicate findings in a way that is useful for product, engineering, and business stakeholders.
Michaël Pilaeten
SOFICO, BelgiumBreaking the system, helping to rebuild it, and providing advice and guidance on how to avoid problems. That's me in a nutshell. With over 20 years of experience in software consultancy in a variety of environments, I have seen the best (and worst) in software development. I'm responsible for guiding consultants, partners, and customers on their personal and professional path towards excellence. I'm chair of the ISTQB Advanced work group, author and international keynote speaker.
ISTQB Agile Tester 2.0: a story on not-so-agile development
In May 2026, ISTQB launched the second version of the ISTQB Agile Tester syllabus.
This session will tackle the why, how, who, when and where of the new syllabus.
What's in it for you? What's new? What's the value?
A full insight, including Q&A.
André Verschelling
ALTEN, BelgiumAndré Verschelling is a Practice Manager, Competence Lead and Principal Consultant Testing at ALTEN, and a board member of the BNTQB. He specializes in designing and implementing test strategies and quality processes, particularly within complex and R&D driven environments. His experience covers large, safety critical system development as well as agile software projects requiring fast, high quality delivery.
As a dedicated trainer and coach, André supports the professional growth of testers by delivering ISTQB® based training programs and tailor made courses that focus on practical testing skills and craftsmanship. As an ALTEN Test Ambassador, he actively promotes best practices, innovation, and knowledge sharing within the testing community.
In his role on the BNTQB board, André contributes to strengthening the testing profession in Belgium and the Netherlands by helping shape the regional body of knowledge. As webmaster, he ensures that essential information and resources from the BNTQB and its partners—such as training and exam providers—are accessible to the broader testing community.
Why you need to become a better tester before getting better in testing with AI
The use of Generative AI has become almost as natural for many developers and testers as having a cup of coffee at the start of the workday. AI has evolved into one of the many tools professionals rely on daily. Yet the old saying still applies: “A fool with a tool is still a fool.”
As a result, many organizations are rapidly developing training programs to help employees work with AI responsibly and effectively. These programs range from courses on ethical considerations to trainings focused on prompt engineering—or a combination of both. The ISTQB® Certified Tester Specialist Level – Testing with Generative AI syllabus is a good example of such a combined program, addressing both ethical aspects and prompt engineering.
If we assume that training is intended to enhance your knowledge and skills, it seems logical to conclude that these programs aim to make you a better tester. However, one could argue that we must first become better testers ourselves before we can meaningfully expand our AI related testing skills. The ISTQB syllabus correctly highlights several risks, but to effectively mitigate those risks, testers must be more critical, skilled, and insightful than the AI tool or LLM they are using.
For instance, how can you determine whether a model is hallucinating when applying specific test design techniques, if you do not fully master those techniques yourself? And how can you detect reasoning errors if you can’t evaluate the flow by lack of knowledge or experience? A tester’s true “superpower” lies in the ability to derive the right test conditions and test cases from a complex test basis—with the required coverage and depth to verify that the system is built right, and to validate that the right system has been built.
However, in various training programs—such as the ISTQB® Core Advanced trainings—I often see that many testers struggle especially with the basic concepts, such as risk management and applying test design techniques. The greatest risk in using Generative AI in testing, therefore, is that the tester falls short in assessing the quality and reliability of AI generated results.
In this presentation, I will give examples on why we need even better testers when applying Generative AI. To that purpose, we explore the key risks associated with using Generative AI in testing and demonstrate how testers can leverage their unique superpowers to mitigate these risks effectively. As a result, you will no longer be the fool with the tool, but you’ll be supercool using the tool.
Taia Dimitrova
ASSIST Software, RomaniaTaia is a Senior QA Automation Architect with 10 years of experience and one core belief: great test automation isn't about replacing testers, it's about giving them superpowers.
Armed with Playwright, WebdriverIO, and FlaUI, she has delivered automation solutions across web, desktop, and simulation applications, cutting regression time by 60%. She actively integrates AI coding assistants like GitHub Copilot, Claude, and Augment to accelerate test generation and make debugging feel slightly less like detective work.
Currently, she is building a test automation framework for a 20 year old desktop medical application with 1,500+ manual test cases. Drawing from her ISTQB Advanced Level - Test Automation Engineer certification and years of mentoring QA teams through automation transformations, she's crafting a framework that's built to last. It's part archaeology, part engineering, and entirely the kind of challenge she signed up for.
Outside of work, she runs on strong coffee and recharges at music festivals, because debugging all week requires a proper soundtrack.
From Chaos to Confidence: A Case Study in Legacy Test Automation Modernization
What do you do when you inherit a test automation mess: flaky tests, no clear structure, and a 15+ year-old product with 10+ client configurations? This case study shares the real journey of transforming chaotic test automation into a structured, reliable framework. No sugarcoating, I'll share what we faced, the decisions we made, and what actually worked. This is not a success story with a perfect ending; it's an ongoing transformation with honest lessons about assessing inherited automation, prioritizing when resources are limited, and introducing structure without stopping delivery. A practical talk for anyone who has ever looked at legacy test code and thought, "Where do I even start?"
Almantas Karpavicius
Nord Security, LithuaniaAlmantas is a seasoned software development leader, author, and educator with a decade of experience in the tech industry. Over the past three years, he has successfully transitioned into management, leading engineering teams while staying at the forefront of the AI revolution. Almantas has built a hundred AI agents and several prototypes. He is the author of two books, including a guide on software development with AI. Almantas has founded two programming bootcamps and created over 100 educational videos on YouTube. His contributions to the tech community have been recognized globally with a 3-year Microsoft Most Valuable Professional (MVP) award. Regardless of being a programmer, Almantas had always a huge interest in automated tests and their impact in software development - he has shared this knowledge in several conferences across Europe.
Measuring the Agent: An Experiment-Driven Look at TDD, BDD, and 'No Testing'
The shift toward agentic development has flipped the software engineering bottleneck. We no longer wait for developers to manually type out features; instead, QA and engineering teams are being buried under an avalanche of AI-generated code. When an LLM becomes the primary author of a codebase, traditional testing paradigms aren't just safety nets anymore - they are the constraints that steer the AI.
Almantas pulls back the curtain on the messy reality of agentic workflows by showcasing how they were used to build Arcanum Audio, a complex multi-track mobile audio mixer. Through a series of prototypes, this session moves past AI hype to look at the hard data of AI-driven quality.
In this session, we will look at:
Whether you are a tester wondering how to handle the sudden flood of AI pull requests, an automation engineer looking to adapt your frameworks, or a quality leader trying to scale output without sacrificing stability, you'll walk away with live-fire demo results and a "bruised" reality of testing ai-generated features.
Dimitar Pop-Trpev
Avenga, North MacedoniaMy name is Dimitar Pop-Trpev, a Senior Test Automation Engineer at Avenga with over 18 years of professional experience, including 10+ years in Quality Assurance and Test Automation. My expertise includes designing and implementing automated testing solutions using Java/Selenium, TypeScript/Playwright, and C#/Playwright, as well as working with Jenkins, GitHub, and Azure DevOps to support CI/CD pipelines and modern software delivery practices.
For more than 5 years, I have also been a mentor in the Avenga Academy for Software Testing, where I help aspiring QA professionals build practical skills and grow their careers in software testing.
Outside of work, I enjoy gaming, 3D modeling and 3D printing, and going on road trips to explore new places and experiences.
Building a Scalable Test Automation Strategy in CI-CD
A scalable test automation strategy within a CI/CD pipeline is crucial for preserving software quality while supporting fast and continuous delivery. It requires a balanced test pyramid, modular and maintainable automation frameworks, and efficient practices like parallel execution and cloud-based environments to handle growing test demands. Ultimately, success depends not only on tools but also on strong collaboration across teams, continuous optimization, and integrating testing seamlessly into the development workflow.
Nadia McKay
Nexa IQ, United KingdomNadia McKay is the Founder of Nexa IQ and a Quality & Testing Consultant with over 30 years of experience delivering technology and business change across a variety of industry sectors, startups, and scale-ups.
Having held senior leadership roles including Head of Testing, Nadia now helps organisations build better products by combining quality thinking, practical testing expertise, and a strong focus on people. She is passionate about the intersection of technology and humanity, exploring how emotional intelligence, communication, and curiosity remain critical skills in an increasingly AI-driven world.
A regular speaker, mentor, and trainer, Nadia is known for her engaging storytelling style, ability to make complex topics accessible, and willingness to challenge conventional thinking with warmth and humour. Through her work with founders, product teams, and technology leaders, she advocates for building not only the right products but building them right.
When she's not talking about quality, resilience, and the future of technology, you'll usually find her connecting people, supporting emerging talent, or searching for the next great conversation over coffee.
Emotional Intelligence in the Machine Age: The Human Side of Software Quality
Software Testing and Quality Assurance often sit in the shadows—essential but invisible. This session aims to pull it into the spotlight, exploring how testing has been integral in shaping our lives, from the steam era to AI-driven platforms. Through the lens of the inventors, I will take the audience on a journey through the industrial and social revolutions.
We will delve into the users of tech – ‘us’ aka human beings, aka our customers, how we are affected by our generation groups, and what influenced each group the most. We will have audience interaction leveraging nostalgia, revealing how each generation’s expectations of technology have shifted, how they are impacted, and what that means for great customer experience and resilience.
With a touch of humour, interactive ‘surveys’ and a lot of lived experience, I’ll explore:
Key takeaways:
Vojin Popovic
Svea Bank, SerbiaVojin Popović is the co-founder of Court of You, an AI-assisted psychotherapy and self-reflection platform. He is also Principal Developer at Svea Ekonomi, Serbia, with 19 years of experience in QA and software quality. Although he started his career close to development, testing and quality improvement became his main professional focus. Besides software, he has a strong interest in psychology and psychotherapy.
When Good User Data Goes Bad: Testing Transformations Before AI Training
This presentation draws on practical experience from a self-reflection and psychological skills platform built around repeated daily and weekly inputs. Examples include daily emotional check-ins, balance moment logging, a 28-day Fixed-Role Voice Lab, and a four-week sleep monitoring and habit-building flow. In such systems, user data is not valuable because of a single entry. It becomes valuable because repeated entries over time form behavioral sequences, patterns, and change signals. Those signals later become candidates for reports, summaries, pattern detection, and eventually anonymized input for AI training intended to support predictions and interventions.
The central argument of this paper is that in longitudinal products, quality assurance must defend not only technical correctness but also meaning preservation. A record can be valid in the database, complete in the export, and still be unusable for analytics or AI training because the behavioral meaning was altered during capture, sequencing, transformation, aggregation, or anonymization. This risk is especially high in platforms where inputs are intentionally spread across days or weeks and where later intelligence depends on patterns rather than isolated records.
The talk presents a practical framework for testing these systems. It outlines three reusable takeaways, defines the core risk of “valid data, broken meaning,” proposes a four-layer test strategy for longitudinal products, describes the testability hooks required to make such systems realistically testable, explains how behavioral test data differs from ordinary record-based test data, and shows why pre-release simulation and post-go-live monitoring are both necessary. It concludes by arguing that early AI readiness should be assessed from the perspective of input quality long before any model is evaluated.
Qambar Raza
CGI, United KingdomQambar Raza is a Microsoft MVP in Developer Technologies and Technical Lead for Platform and Reliability Engineering at CGI, where he builds AI-augmented testing infrastructure for software delivery. He spent four years as a Principal Engineer on BBC iPlayer and BBC Sounds, scaling validation platforms across web, mobile, smart TV and set-top box for 500 million monthly users. He is LinkedIn Learning's top Playwright instructor, with over 80,000 engineers trained worldwide. He writes and speaks regularly on AI-augmented testing, Playwright at scale, and the people-first quality mindset.
Testing AI Agents with Playwright and MCP: A Practitioner's Guide
AI agents are no longer theoretical — they browse the web, fill out forms, make decisions, and interact with production systems. But how do you test something that behaves differently every time it runs?
This talk presents a practical approach to testing AI agents using Playwright and the Model Context Protocol (MCP). Drawing on real-world experience building AI-augmented test frameworks and training 54,000+ engineers through LinkedIn Learning, I will demonstrate how to combine Playwright's browser automation capabilities with MCP to create reliable, repeatable tests for AI-driven applications.
We will cover: setting up Playwright to observe and validate AI agent behaviour in the browser; using MCP as a bridge between your test framework and AI models; strategies for handling non-deterministic outputs — when your test subject does not do the same thing twice; building confidence metrics instead of pass/fail assertions; and real patterns from production systems where AI agents interact with users.
Attendees will leave with a working mental model for testing AI agents, concrete code examples they can adapt to their own projects, and an understanding of where traditional test automation ends and AI-specific testing strategies begin. This is not a theoretical overview — it is a practitioner's toolkit for a problem most teams are already facing.
Elena Kulgavaya
Source2Sea, PolandElena is a Lead QA Engineer at Source2Sea with over 15 years of experience in software quality, test automation, and delivery engineering. She specializes in contract testing and helping teams build reliable, maintainable test strategies for distributed systems and microservice architectures. Elena is the creator of Surety, an open-source contract testing framework, and writes about practical QA strategies at qaexplained.com.
From E2E to Contract Tests: A Practical Migration Path Beyond Microservices
End-to-end tests are a natural starting point. But as system grows, they become slow, flaky and expensive to maintain. Contract testing offers a way to regain fast feedback and stability, but it is often treated as a microservices-only technique, focused on isolated service interactions.
In practice, many critical contracts exist at a higher level - between the frontend and the backend, where real user interactions happen. Yet these interactions are typically only verified through end-to-end tests.
In this talk, I'll present a practical, step-by-step migration approach based on real-world experience. Starting from existing E2E scenarios, we replace backend interactions with mocks, then use those interactions to build contract tests and progressively extend coverage across services.
The approach allows teams to decompose large end-to-end tests into smaller, faster, and more reliable checks, while preserving the confidence those tests originally provided and expanding contract testing beyond service boundaries.
We'll also explore the real cost of this migration: what is straightforward, where teams encounter friction, and what trade-offs to expect in different stacks and environments.
You'll leave with concrete migration strategy and a clearer understanding of how contract testing can evolve beyond microservices without starting from scratch.
Oleksii Bakunin
EDGE, UkraineQA Engineer with 14 years of experience in software quality engineering. Has built QA organizations, designed automation frameworks, and led quality initiatives for startups and enterprise products across FinTech, EdTech, Travel, VoIP and more.
In recent years, was focused on transforming QA with AI -- developing AI-powered testing assistants, improving engineering workflows, and exploring how generative AI is changing the role of quality engineers. I'm enjoying sharing practical lessons learned from real production environments and helping teams prepare for the future of software testing.
AI-Augmented QA in Practice: How to Build a Controlled LLM Workflow
LLMs can generate test cases, summarize requirements, review documentation, write automation and much more. Yet many teams discover the same problem after the initial excitement: the more AI is involved, the harder it becomes to understand, review, and trust the outcome.
This session presents a practical approach to AI-augmented Quality Engineering based on controlled AI workflows—where AI accelerates repetitive work, while engineers retain ownership of every critical decision. Instead of replacing QA processes with autonomous agents, we build deterministic workflows with clear system boundaries, approval gates, and measurable outputs.
Drawing on real production experience, you'll learn how to:
Based on real production experience, this session demonstrates how to scale QA impact while maintaining reliability and control.
Verica Solin
IWConnect, SerbiaI'm a Senior QA Consultant at IWConnect with over 15 years of experience in software quality engineering across e-commerce, insurance, IoT, and telecom. I specialize in test automation, AI-assisted testing, and quality strategy, helping teams build reliable software and deliver products that create real value for users. Having worked in both QA leadership and Product Management roles, I enjoy bridging the gap between quality, technology, and business goals. I'm passionate about the future of QA and love sharing practical insights on how AI is reshaping the way we test, collaborate, and deliver software.
Evaluation Prompts: The future of QA
In the era of AI, LLM-based applications produce probabilistic outputs, making conventional testing approaches—based on fixed inputs and expected outputs—insufficient.
This shift introduces a fundamental challenge for Quality Assurance: how do we validate systems where correctness is subjective, context-dependent, and non-deterministic?
Thus, QA needs to transform, from deterministic to probabilistic evaluation. This presentation proposes evaluation prompts as a practical and scalable solution. By leveraging LLMs to evaluate other LLM outputs, QA teams can move from rigid assertions to structured, multi-dimensional evaluation. This approach enables systematic testing of correctness, consistency, reasoning quality, safety, and business relevance—positioning QA as a critical function in the era of AI-driven systems.
