Stop Tuning Hyperparameters. Start Looking at Your Data.
Galia Kraicheva
My journey into statistics began in 2006 at my university in France, where I discovered the beauty behind data and modelling.
Ten years later, my love for Ruby sparked through the Rails Girls study groups in Sofia — and I’ve been part of the Ruby community ever since. From attending meetups and being a Rails Girls Instructor to being an MC at Balkan Ruby, this community has shaped a big part of my life.
I’m excited to be back on stage — this time as a speaker — bringing together my two worlds: Ruby and data, in a conversation about AI and the mistakes we make when we forget to truly understand our data.
About
Why do so many AI projects fail — even with fancy models and impressive metrics?
Because we obsess over algorithms and ignore the one thing that actually matters: the data.
In this talk, we’ll challenge the “model-first” mindset that dominates modern data science. Instead of jumping straight to XGBoost, deep learning, or the latest AI trend, we’ll step back and ask a simpler question: Have we actually looked at our data?
We’ll talk about how skipping proper exploratory data analysis leads to false confidence, hidden leakage, biased predictions, and systems that work great in a notebook… and fall apart in production. Through practical examples and clear principles, I’ll argue that understanding your data-generating process is more valuable than any hyperparameter search strategy.
Better models don’t start with better tuning. They start with better questions.
Sponsors
By sponsoring Balkan Ruby, you are helping us make a great event while promoting your brand to the passionate Ruby developers in Bulgaria, the Balkans, and beyond!
Tickets
Support Balkan Ruby 2026 by purchasing a ticket. ❤️