Eduardo Rdgz-Á

What's lived beats what's designed.

The dominant AI discourse in 2026 is still centered on the model: how big, how well benchmarked, how fast. But the AI products that deliver real value share something that never shows up in benchmarks: behind them is someone who knows the problem because they lived it, not because they researched it.

That conviction organizes my work. I come from three places — physics, eight years in the classroom, and building and operating an AI-native platform in production — and all three tell me the same thing: the sustainable advantage lies in the distance between whoever designs the solution and whoever suffered the problem. The shorter that distance, the better the product.

What this means in practice:

• — An unoperated model is not a product. Governance doesn't live in a document: it lives at runtime — in the log, in the queue, in the versioned prompt, in who signs off on every output.
• — Most of what we call “educational problems” is bureaucracy stealing time from people trying to do their jobs. I learned it in the classroom; I confirmed it building ILC, where a user told me: “We have global coherence across all 250 pages of lesson planning.” The system works because it solves a problem I lived, not one I imagined.
• — The harness is human. Sustaining long chains of thought for hours — where the real breakthroughs appear — is human work. Responsibility for what the system generates is not delegated to the algorithm.

I'm drawn to teams building real DS/ML products with AI in production — with accountability anchored in a person, not in the algorithm.

Full resume →