Shipping agent graphs without burning your margin budget
Autonomous agents are cheap to demo and expensive to keep honest. The moment you wire “just one more tool” without a contract, your margin graph looks like a crypto wallet after a bridge hack.
AI Director · Systems architecture
Agentic AI, shipped.
AI Director architecting agent fleets, LLM infra, and RAG in production. Madrid · EU-remote.
System status
shipping · agents v3
building · startlight-brain-app (semantic search + pgvector)
writing · 04 essays in flight
speaking · two keynotes booked
Selected work
Three pieces from the field — agentic platforms, retrieval infra, and an applied transformer pipeline.
01
Agentic knowledge acquisition studio
Internal control plane tying multi-agent task graphs, retrieval, and evaluator hooks for regulated dossiers (stub brief).
Cold-start agent swarm <11 min median (staging) Agentic platforms 2025 02 startlight-brain-app
A "second brain" app with semantic search, pgvector embeddings, knowledge graph, and Web Clipper — built end-to-end as a solo project.
45 commits · semantic search pipeline with hybrid retrieval (cosine + top-K) Personal project · Agentic RAG 2026 03 Enterprise grade hybrid RAG gateway
Opinionated ingestion + retrieval façade unifying lexical + dense signals for marketplace scale (stub brief).
Rolling index rebuild P95 6m12s vs 54m baseline Retrieval infrastructure 2024
Internal control plane tying multi-agent task graphs, retrieval, and evaluator hooks for regulated dossiers (stub brief).
Cold-start agent swarm <11 min median (staging) Agentic platforms 2025 02 startlight-brain-app
A "second brain" app with semantic search, pgvector embeddings, knowledge graph, and Web Clipper — built end-to-end as a solo project.
45 commits · semantic search pipeline with hybrid retrieval (cosine + top-K) Personal project · Agentic RAG 2026 03 Enterprise grade hybrid RAG gateway
Opinionated ingestion + retrieval façade unifying lexical + dense signals for marketplace scale (stub brief).
Rolling index rebuild P95 6m12s vs 54m baseline Retrieval infrastructure 2024
Selected press & long-form
Where the work and writing have shown up over the years.
- La Razón2019
- Amloii (YouTube)ongoing
- Editorial2016
- Mediumongoing
Engagements
How teams use my time when an essay isn't enough — selectively, with prep.
Advisory
Sparring with CTOs and Heads of AI on architecture, evals, and team shape.
Implementation
Hands-on 0→1 work when a deck won't get the system out the door.
Speaking
Keynotes, panels, and editorial briefings on agentic AI in production.