I'm Néstor Martínez.

I started as an Industrial Engineer with a Master's in Business, five years into technology consulting before I noticed the pattern: the AI projects that mattered weren't the ones with the most impressive demo — they were the ones still running, unattended, six months later.

For the last two years I've worked exclusively on production AI systems: multi-agent architectures that read documents, answer on WhatsApp, draft emails, and hand off cleanly to a human when they should. Seven to eight client projects delivered, from a two-partner law practice to a 248-advisor financial network.

I care less about the model and more about what survives contact with real users — human-in-the-loop learning, approval gates before anything ships, cost-aware routing so the frontier model only runs where it earns its price. A RAG assistant that has to quote a source word-for-word and a CRM copilot for insurance advisors need very different engineering, but the same discipline.

What keeps me in this work isn't the architecture diagrams — it's watching a two-person law firm get seven hours a week back, or a team that ran its entire operation on spreadsheets get its first real system. The numbers are the proof; the time recovered is the point.

If you're evaluating for a role, or have a system that should be running and isn't — I'm easy to reach.