APEXiA

Engineering, in detail

How we build and run APEXiA: compound AI, self-governing AutoML, catalogs that can't rot, and operations research on the street.

2026-06-12 DBA · ES

The data catalog that can't rot

A data dictionary verified against the ERP's own source code, a semantic join layer, and a live database map — maintained by agents, not by good intentions.

2026-06-12 OR · ES

Routes that optimize themselves: operations research on the street

TSP for the rep's visits, CVRP for truck dispatch — with the real fleet, real weights, and the real price of diesel.

2026-05-28 Data science · ES/EN

The self-governing forecast

An AutoML system that proposes, evaluates and prunes its own features — and why we let it govern itself.

2026-05-20 Data science · EN

Why a compound AI system beats purely agentic AutoML

Determinism where it matters, LLMs where they help: the architecture behind our production models.

2026-05-15 Architecture · EN

The compound AI architecture that replaces an entire data team

The technical whitepaper: how APExIA chains classification, SQL, models and orchestration over a canonical data layer.