AI Product Manager · Barcelona
I take LLM systems from demo to production — and prove they work.
8+ years product across B2B and B2C SaaS. Now owning a multi-agent platform with RAG, evals and human-in-the-loop in a regulated, high-liability domain.
8+
years in product
~30%
messages AI-drafted
1k+
users shipped solo
4
case studies
Selected work
All work →ORCHESTRATOR → 3 AGENTS
Settly · 2024–Present
Relocation AI Agent Platform
A multi-agent system that turns relocation expert knowledge into platform capability — retrieval, evaluation and human-in-the-loop.
MULTI-AGENTRAGLLM EVALUATIONHITL
Read case study →SETTLY → VENDOR → CAPTURED
Settly · 2024–Present
Vendor Communication Platform
Centralised vendor communication that cut response times in half and ended inbox fragmentation.
PLATFORMWORKFLOW
Read case study →3× FASTER SCHEDULING
Rentman · 2022–2023
Product Vision for Crew Scheduling
Repositioned a mid-market scheduling tool as an enterprise operations backbone — 80% revenue growth.
VISIONJTBD
Read case study →0–8H
8–10H
10–12H
12H+
Rentman · 2022–2023
Custom Pay Rates
A flexible rate engine that unlocked U.S. market entry — 25% more license upgrades.
PRICINGU.S. EXPANSION
Read case study →How it is measured
Golden set · n=420
Retrieval recall0.91
Relevancy (judge)0.86
Edit distance (inv.)0.74
Normalized edit distance runs on every interaction — the passive quality signal that needs no reviewer.
“Normalized edit distance is a passive quality signal on every single interaction — no reviewer required.”
Stack
Pydantic AILangfusepgvectorCohere RerankDeepEval
Track record
2026–
Studio Firn
Solo consumer products
2024–26
Senior PM — Settly
AI agent platform, immigration knowledge work
2017–23
Rentman
Specialist → Senior PO, B2B ERP