About this Event
Two talks, one conversation, and a room full of developers figuring out what’s worth automating.
AI agents promise to take the boring stuff off your plate. But what does it take to make them work? And when does adding more tooling just create more work?
Join us in Barcelona for an evening of practical lessons, hard-earned insights, and meaningful connections. Semrush engineers, Daria Naidikova and Nick Kalion, will share what they learned building and testing AI agents — from breaking complex tasks into reusable skills to finding out when leaving things alone delivers better results.
*Please note that there will be photos and videos taken at the event. If you do not wish to be recorded or photographed, reach out to the event host on the spot.
**To enter the building, you'll need to show an official ID (passport or NIE/DNI) to security. Be sure to provide your full name when signing up.
Agenda
🕑: 06:00 PM - 06:30 PM
Doors open
Info: Welcome drinks and snacks
🕑: 06:30 PM - 07:00 PM
Icebreaker game
Info: Prize included!
🕑: 07:00 PM - 07:30 PM
Boring task, hard lessons: Building an AI agent that works
Host: Daria Naidikova, Frontend Developer
Info: Getting an agent to do a task is one thing. Getting it to do that task reliably is another. Daria shares the hard lessons behind building an agent that works — and packaging it so other teams can use it too.
What you’ll take away:
Why agents need reference docs, not just examples
How to split complex work into small, testable skills and coordinate them with an orchestrator
Why output validation belongs in a separate, isolated step
How to package your solution for reuse across teams
🕑: 07:30 PM - 08:00 PM
Doing nothing worked: What I learned testing AI agents across multiple reposit
Host: Nick Kalion, Senior QA Engineer
Info: More tooling. More context. Better results? Not always. Nick shares what happened when he tested AI agents across multiple repositories — and why “do nothing” deserves a place in your experiments.
What you’ll take away:
Why you should test added tooling against a “do nothing” baseline
How measuring actual failures helps you decide what needs fixing
How extra context can lead to noise and unnecessary edits
Why it helps to keep only the facts agents can’t infer from the code
🕑: 08:00 PM - 09:00 PM
Closing remarks, drinks & networking
Event venue & nearby stays
Semrush Barcelona Office 6th Floor, 98 Carrer de Tànger, Barcelona, Spain
Concerts, fests, parties, meetups — all the happenings, one place.