At NuMind, we build machine learning tools for information extraction using custom-trained language models. But we’re not here to talk about the models—we’re here to talk about the full-stack Scala platform that makes working with those models possible.

In this talk, we’ll share our experience building a full production system entirely in Scala—from backend to frontend—using a stack that includes Ox, Tapir, JOOQ, Laminar, Keycloak, Shoelace, and Vite. By integrating these tools into a cohesive, type-safe environment, we’ve created an ecosystem that is both developer-friendly and production-ready—allowing us to move fast without sacrificing consistency or robustness.

What makes this journey even more interesting is that one of us started as a Scala beginner—and was able to contribute to backend code from day one. This speaks volumes about the approachability and consistency of the ecosystem we’ve assembled.

We’ll talk about:

• Why we chose Scala for full-stack development
• The architectural patterns that worked for us (and the ones that didn’t)
• The challenges we faced
• And most importantly, how this stack empowered our small team to move fast.

If you’re curious about full-stack development in Scala, or want inspiration for building real-world ML platforms, we’ll show you what’s possible when the whole stack speaks the same language.