Sergio Barón

I design automation systems for businesses by mapping where a process breaks, deciding what the system should do about it, and building it end to end. I focus on the decisions behind an AI system—not just a single tool—so the logic is auditable and failures are part of the improvement loop.

My most recent project is a campaign monitoring tool that detects anomalies in paid media accounts and uses an LLM to explain the likely cause. I built it with reliability in mind: when the model generated confident explanations that the data couldn’t support, I removed those cases from the model’s reach instead of forcing the prompt. I also specified how findings are identified to prevent duplicate alerts and verified behavior by rerunning the full batch twice on the same data.

Along the way, I document design mistakes and fixes publicly so teams can see the reasoning, not just the outcome. I also build AI agent workflows—like a WhatsApp customer service agent that routes messages through Twilio, interprets them with an LLM, and answers automatically—using reusable n8n templates and clear architecture decisions.