Engineering AI
that ships.
I design, ship and operate production GenAI systems for regulated environments — RAG pipelines, multi-agent orchestration, enterprise integrations — with the engineering rigor that compliance demands.
Madrid, Spain · from August 2026 · EN / FR / PT / ES
Tech-first, end-to-end.
My professional path is built on engineering — production AI work today, stacked on real software experience and a technology degree underneath.
AI Solutions Consultant → AI Engineer
Joined as an AI solutions consultant and was promoted to AI Engineer. I designed and shipped production GenAI systems in regulated tax and compliance environments, where traceability, DLP and auditability are the entry ticket — governance is part of the architecture.
- Built a legal/tax semantic search engine with 94% validated accuracy using agentic orchestration
- 30–80% execution-time reductions across AI and automation initiatives
- 1,600+ professionals trained on responsible AI adoption
Co-founder & AI Engineer
The academic-services company where I am the entire engineering department. I built our production multi-agent review platform end-to-end — Django, PostgreSQL/pgvector, Celery, Docker — with Langfuse tracing and Sentry in production. Nothing teaches production reliability like operating a system your household depends on.
- Cut a six-hour expert review task to fifteen minutes with a 5-pillar multi-agent system
- Built the funnels and landing pages behind ~90% monthly revenue growth in one quarter
- Full production observability: Langfuse tracing, Sentry, automated health monitoring
Full Stack Developer
Built the company's digital foundation end-to-end — its first institutional website, CRM integrations, and operational automation for commercial and administrative processes.
- Developed the company's first institutional website and digital-presence strategy
- Automated quotation generation and inventory-control workflows
- Built customer-service automation supporting 120%+ revenue growth in the period
Technologist Degree — completed
Full technology degree covering software engineering, systems architecture, databases, and applied AI — completed with a final GPA of 8.8/10. The theoretical foundation under the practical work.
- Programming logic, data structures, and algorithms
- Database design, SQL, and systems architecture
- Web development, APIs, and applied AI
Building production AI is engineering work — not prompting.
The shift from "interesting prototype" to "system that runs in production for regulated workflows" lives entirely in the engineering: data pipelines that don't break, retrieval that's auditable, agents whose handoffs are traceable, validation layers that catch what the model gets wrong, and governance baked in instead of bolted on.
That's why I lean on the engineering background. A production multi-agent platform I co-founded and operate, a 3-person Unity team I led, a systems degree finished with an 8.8/10 GPA — none of it is glamorous, but it's where the discipline comes from. The AI part is what gets the headlines; the engineering is what makes it ship.
I'm driven by depth: when something interests me, I go all the way in. That's how I graduated top of my cohort at the University of São Paulo (9.2/10 GPA) and earned a merit scholarship to Université Lumière Lyon 2 in France. When I chose technology, I brought the same standard — a full systems degree finished with an 8.8/10 GPA, a promotion at KPMG, and a peer-reviewed paper on multi-agent system architecture — the same architecture I run in production. Different fields, same pattern.
Four principles I won't compromise on.
Production-first
I architect for the live system, not the demo. The work that survives contact with real users is the work that matters.
Ship to learn
Prototypes are cheap; running systems generate signal. The fastest path to a great system is a deployed one — then iterate fast.
Multi-agent by default
Decompose complex reasoning into specialized agents with clear contracts. Easier to test, debug, evolve, and explain.
Build for adoption
A deployed system nobody uses is a failed system. Enablement ships alongside the code, not as a phase after it.