DevFest Kraków 2026

Official opening of DevFest Kraków 2026, welcome remarks, and event overview.

For the past 30 years we've been building websites for humans, now your website has a new type of visitor - AI Agent. How agents can perceive your website? What are WebMCP tools and how can they help? What interactions channels can we expose to agents? How to test integrations? Cool demos included :)

I'm a Product Owner. I'm the one who sometimes calls on Friday at 5:45pm asking "can you stay a bit longer tonight?" I'm the one pushing for the deadline. I'm the one who sometimes has to hear you explain why something didn't work. I know this table from both sides, and I know exactly which of your moves actually work, and which ones just confirm to me that I can keep pushing.

This talk is three scenes that play out between you and someone like me, each one with a concrete technique to take back to the office on Monday:

Say No — how to assertively refuse overtime when you hear "it's just this once," without burning bridges and without the guilt trip.

Convince — how to talk a PM or client into moving a deadline when the official answer is "there's no such option," by showing them it's actually in their own interest.

Apologize — how to explain a screw-up so the team and the client still trust you afterward, instead of starting a hunt for someone to blame.

No management-book theory. Concrete negotiation and communication techniques I know from the other side of the table — and that actually work once you use them on purpose.

Description coming soon.

In this live building session, I will create a production-ready, event-driven AI pipeline on Google Cloud from a blank terminal to a working system.

A file lands in Cloud Storage, Eventarc routes the event, Cloud Run processes it reliably, and Vertex AI with Gemini enriches the data through classification, summarization, and entity extraction. The enriched results are then stored.

The focus is not on AI hype or a pre-recorded happy-path demo. Attendees will see the system being built live, including the practical decisions, trade-offs, and failure scenarios involved in making it production-ready.

Two years ago the recipe for a company AI assistant was simple: dump your documents in a database, find similar chunks, paste them into the question. Today we know why that fails so often.

Three things: why similarity search alone has a hard ceiling, why a model would rather invent an answer than say "I don't know", and what today's systems look like when they can search many sources at once and check their own work.

For anyone who already has a RAG in production, or is about to pay someone to build one.

For a decade, Android developers built for a 6-inch phone screen. Today, Android powers smartwatches, TV living rooms, and connected cars. In this session, learn how to adapt your engineering mindset across three physical environments—Wear OS, Android TV, and Android Auto—and architect a clean, modular codebase with shared Kotlin domain logic that runs seamlessly across the entire Android ecosystem.

AI is great at generating code, but terrible at understanding what’s actually happening inside your browser. We still copy-paste logs, screenshots, and half-formed thoughts into chat windows and hope for the best.

Model Context Protocol (MCP) fixes this by letting AI consume real, structured context directly from Chrome DevTools. Instead of guessing, AI can see network requests, performance traces, console output, and runtime state — just like a human developer.

This talk introduces MCP without the buzzwords and shows why DevTools are the perfect place for it. We’ll walk through concrete debugging and performance scenarios where MCP turns AI into a collaborative partner, not an autocomplete engine. Expect nerdy details, practical demos, and a few “why didn’t we have this earlier?” moments.

Description coming soon.

Inspired by a friend of mine, I started to purchase domains, new ones with clean page rank. Sounds cool, but what to do with those? I started to build a system around Astro (and I am nextjs boi), Cloudflare. First suprise - indexers gives a f**k about my AI-generated page. Who would have though.... Let me show you what I started to connect (subagents, skills, MCP, APIs) so that we stop producing AI slop websites.

Integrating AI into modern applications usually comes with steep API bills, latency, and compliance headaches. The fix might already be running on your user’s machine.

Welcome to Web AI. This session breaks down how running Small Language Models (SLMs) and leveraging native browser APIs makes app intelligence instant, offline-capable, and completely private. Learn how client-side inference eliminates third-party API costs while delivering smarter, privacy-by-design web experiences.

Join us for an evening of networking and learning from key industry leaders. You’ll hear how they navigated real-life career challenges, learn what worked (and what didn’t), and get your questions answered in real time.

Mentors & Leaders:
• Tomasz Manugiewicz — evoke, General Manager & Author
• Katarzyna Bukowska — FedEx Dataworks, Director of Engineering
• Magda Rabiej — Google Krakow, Head of IT Development
• Krystyna Pietrzykowska — Hitachi Energy, Head of Technology Center
• Łukasz Domoradzki — GE Healthcare, Senior IT Director
• Andrei Sheleh — Splunk, Director of Engineering

Cavendish Professionals’ TECH TALK: Live Mentoring Sessions bring tech enthusiasts and industry leaders together to inspire growth, learning and connection. The sessions offer a unique opportunity to gain career guidance and insights directly from seasoned professionals, helping attendees develop the skills and confidence to become future leaders in tech.

Description coming soon.

Official event wrap-up, community recognitions, prize giveaways, and closing remarks.