Join us in Iasi in 2025!

This keynote uncovers a new way to think about architecture—as code. Architecture as Code is a new concept that allows you to describe an architecture through executable source code, therefore allowing you to govern the architecture as well. In this session we describe numerous intersections of software architecture with the organization, defining each intersection using architecture-as-code to verify that the architecture is properly aligned.

It’s never been a stranger time to try to learn and practise TDD than now: the core of programming is changing, demands for productivity have never been higher, and trends never stop changing. More than ever, programmers are desperate for more certainty regarding what they should learn, how it will feel to try, and whether it will be worth the effort. And then there’s Generative AI making the whole situation cloudier. In this talk, I’d like to help you understand what’s likely to actually happen to people and organizations who learn and practise TDD, including how to get through the early difficulties to get to “the good stuff”. I’ll also talk about what programming looks like for people experienced in TDD. (Hint: There’s a lot less “TDD” than you might expect.) Practitioners can feel more comfortable with what they’re likely to go through, while managers and stakeholders can feel less surprised by what would happen to their projects if they let TDD emerge as a central practice. I’d even like to relay a few nuggets of emerging wisdom regarding the unexpected importance of TDD (and Evolutionary Design more generally) in the age of programming with LLMs. No dogma, no value judgments, no exaggerated claims. Old Man, but no Yelling At Clouds. I promise.

What happens when your carefully crafted models collide with messy data, shifting goals, and real-world unpredictability? In this talk, I’ll share my journey of building crypto price prediction systems with traditional ML and LLMs — the roadblocks I hit, the mistakes I made, and the solutions that actually worked. While my project is about crypto, the lessons apply to anyone tackling prediction, analysis, or just the beautiful chaos of machine learning. 

Modern AI systems are only as good as the data they rely on. Yet most machine learning failures don’t come from the model itself — they stem from broken pipelines, silent schema changes, and unvalidated inputs. In this talk, we’ll explore how Data Contracts bring reliability and governance to the modern data stack, using Delta Live Tables expectations in Databricks to enforce quality rules at scale.

With reliable data foundations in place, we’ll shift gears to the exciting frontier of real-time AI. You’ll see how streaming time-series data (like stock prices) can feed into a Databricks pipeline, power an ML model, and even be summarized by a Generative AI layer that produces human-readable insights in real time.

By the end, you’ll understand how to combine governance and innovation: first making data trustworthy, then unlocking AI systems that act — and explain themselves — in milliseconds.

Internal Developer Platforms (IDPs) are quickly becoming the backbone of modern software engineering. Think of them as a one-stop shop where developers can build, test, and deploy applications with speed and confidence. In this talk, we’ll explore why IDPs are on track to become an industry standard—Gartner predicts that by 2026, 80% of engineering organizations will have platform teams offering reusable tools and services. Join this session to learn more about our journey building VOIS internal IDP, and how it’s transforming the developer experience from the inside out.
Whether you’re a developer, DevOps engineer, or product owner, this session will show you how platform engineering is shaping the future of software delivery.

Modern architectures must do more than just store and process data – it’s about enabling real-time, governed, and meaningful access to data across teams and domain. This talk explores how AWS-native Lakehouse architectures – built with S3, Glue, Iceberg-backed S3 Tables, and Bedrock, enable scalable, serverless, and semantically rich platforms for data products. We’ll show how adopting Data Mesh principles – domain ownership, self-service access, and federated governance unlocks delivery models where both technical and non-technical teams can build apps, analytics, and AI workflows on top of trusted data. As a key enabler, we introduce the idea of context engineering: structuring metadata, semantics, and access paths to bridge the gap between raw datasets and LLM-driven insights. This is how modern data platforms move from infrastructure to impact.

If trends are to be believed, modular monoliths are the new kid on the architecture block. They’re sold as an antidote to the complexity associated with overdosing on microservices. Except for one problem: modular monoliths are not a new idea or architectural style, just a new turn of phrase. The current trend is the pendulum swinging back with the benefit of some rebranding. At best it’s “Monoliths, but we’re going to do them right this time, we promise!” At worst it’s “Oh no, here we go again.”

Many teams adopted microservices not for their runtime or deployment benefits, but because they sought refuge from the tangled mess of their monolithic codebases. They wanted microservices because they reinforced partitioning. But as Simon Brown notes, “If you can’t build a monolith, what makes you think microservices are the answer?” In practice, what many teams discovered is that what gets reinforced is poor partitioning.

There’s a generation of developers who have only worked with microservices over their careers, for whom modularity in other forms is a new-to-them discovery. In this talk, we will look at the history of modularity in programming languages and software architecture from the 1960s to the present, taking in the sights and sounds of information hiding, data abstraction, component-based development, process-based architectures and more.

AI agents capable of acting autonomously represent the next step on the road to building intelligent systems that automate decision-making and workflows. Moreover, agents can be combined to form “teams” that collaborate with each other to accomplish complex tasks. Open-source frameworks such as Agno and Semantic Kernel simplify the task of building agentic systems, offer a variety of tools to power autonomous agents, and allow you to write custom tools of your own. Many provide seamless support for tools imported through the Model Context Protocol (MCP) as well. Learn what AI agents are, how to build them using popular agentic frameworks, and how to use them to solve real-world business problems. And see first-hand why agentic AI is a game-changer in the world of generative AI.

3D started with meshes — triangles stacked into models. Then came photogrammetry, pushing realism further. But those methods already feel… so 2022. Today, Gaussian Splatting offers a radical new approach: scenes made of millions of splats, rendered fast and looking breathtakingly real. In this talk, we’ll explain how splats work, how we can practically use them, we’ll explore where they will be in a few years, and why they might just change the future of 3D forever.

Software vulnerability is a huge concern. What’s lurking in code is a question that keeps passionate programmers up at night. Is there a memory leak, what about a race condition, oh what about security issues, are we violating purity of functions when we’re not supposed to? We have to maintain code that others have written and it’s not always easy and quick to detect those defects ticking away in the code. In this presentation we will use AI based tools to detect issues in code, using multiple examples, and apply automated fixes and will reason about our approach and the change.

Legacy systems are often double-edged: They contain years of accumulated business knowledge and value and they enable to pay salaries at the end. But at the same time, they can feel like a burden, slowing down innovation and making organizations less adaptive to change. 

A full rewrite or big-bang reorganization is rarely an option. So how do we evolve a legacy system incrementally without breaking what works? 

This talk explores a pragmatic approach to optimizing legacy systems for flow. It introduces Architecture for Flow, a structured yet flexible approach that integrates Wardley Mapping, Domain-Driven Design (DDD), and Team Topologies. This approach focuses on identifying suitable streams of change, modularizing step by step, and evolving teams alongside the system; ensuring sustainable, low-risk transition from the status quo to an envisioned future.

The many mistakes which destroy teams, sometimes from the outside, sometimes from the inside. And one very specific template for how to build a great team. 

Real talk and examples, all inspired from reality. Let’s have a laugh, shed a tear for all the teams broken before their time, and learn how to do it better. 

With Generative AI increasing being used for software development, developers are pushed towards other roles, including software architecture. The problem is, software architecture is a massive multidisciplinary subject with no clear definitions or job description, making it difficult to chart a course for undertaking all of the learning required to become an effective software architect. That’s where this session comes in. In this talk, Hands-on Software Architect and Author Mark Richards will show you how to develop your skills to start transitioning into a software architect role, and what it takes to become an effective software architect in your organization.