01 / Research line
Applied artificial intelligence
We start applied AI with a concrete decision, the data available to study it and the person accountable for its use. Urban A.I concerns territory, mobility and services; healthcare brings different clinical and operational questions. An exploratory model should never be presented as an automatic decision.
The book La Máquina del Vacío provides a critical frame for this line: what is lost when interactions become patterns, metrics or predictions? Technical research must document sources, assumptions, affected groups and the situations that call for human review.
Three areas of work
Models with purpose
We establish an understandable baseline before testing more complex models and compare error rates across groups.
Supervised agents
Agents consult identifiable sources, leave an audit trail and request approval for sensitive actions.
Continuous evaluation
We test adversarial prompts, missing data and changing context; plausible answers still require expert validation.
Related cases