Luis Miguel Urdaneta Acedo

I'm a civil engineer who builds AI systems for the architecture, engineering and construction industry — and I sit on both sides of that gap. I read structures and construction logic like an engineer, and I ship the Python tooling that makes a design team measurably faster. My current work is multi-agent LLM systems running over real building models: a seven-agent pipeline, built as my Master's thesis at Zigurat, that reads an IFC model through IfcOpenShell and verifies it against acoustic regulations end-to-end — Claude API agents plus deterministic Python modules, with mandatory human-in-the-loop validation and per-agent model routing for cost control. Alongside it I've built tool-calling agents with Agno served over FastAPI, a production RAG layer with vector retrieval and an MCP server, and an unattended agent fleet that has run on a daily schedule for months across roughly 12,000 lines of Python. Before AI, I was a field engineer. Structural design at PDVSA (Petróleos de Venezuela) on offshore oil production platforms — steel access platforms, walkways and equipment-support structures analysed in STAAD.Pro, with the full AutoCAD fabrication package issued straight to the workshop — and site supervision on Metro de Maracaibo's mass prefabricated housing programme, controlling erection of precast elements against drawings. That's where the domain knowledge comes from; it isn't read off a spec sheet. From 2023 to 2024 I worked as an independent AI consultant on internationally funded development programmes, applying LLMs and NLP across large multilingual document corpora in English, French and Portuguese. I'm looking for a fully remote role — employee or contract — with a company building AEC technology, BIM automation, proptech or document-intelligence products. Native Spanish, fluent Portuguese and English, based in Portugal (GMT) with a solid afternoon overlap with US Eastern hours.