Somia’s co-founder was invited as a speaker at the renowned Lean ORP 2025 Congress in Rosario, Argentina. Her talk focused on a powerful message often overlooked in the world of AI: there are people behind every algorithm.

Instead of talking about AI as a mysterious black box, she walked the audience through how it really works—step by step, with clarity and purpose.

Using the well-known Iris Dataset, she explained what features are and how vector transformations operate—one of the foundational concepts behind all AI algorithms, from traditional models to modern transformers.

To make it tangible, she demonstrated how this concept is applied in a Support Vector Machine (SVM) example, helping the audience visualize how data is transformed and classified in multidimensional space.

And here comes the key insight: Transformers do the same thing—but they learn to do it by themselves.

Instead of hand-crafted kernels like in SVMs, Transformers use deep learning to learn these transformations automatically.

This is exactly what attention does: it’s a vector transformation based on the relationship between words in a sentence.

For example:

“I went to the bank” vs “I swam to the bank of the river”

The word “bank” is the same, but the context changes everything.

In the first case, you’d expect the next sentence to be “…and the bank was crowded.”

In the second, something like “…and he was wet.”

Attention in a Transformer captures this.

Each word influences the meaning of others, and this influence is learned. It’s like an attention kernel—trained from data, just like a powerful version of an SVM kernel.

Her intervention was not just technical—it was a reminder that AI is shaped by human intention, expertise, and strategy, especially when applied to crucial fields like occupational risk prevention.

We’re proud to see Somia at the forefront of these conversations, pushing for AI that’s transparent, strategic, and human-centered.