AI and automation that actually work, every day.
LLM agents, integrations and custom software for businesses. In production, with judgment.
Real solutions, in production
WhatsApp AI agent that operates an ERP
Problem: executives had to log into Odoo for every query or change, which was slow and impractical day to day.
Solution: a conversational WhatsApp agent. An LLM interprets each instruction, chooses among several Odoo-connected tools (function calling) and executes the action, asking for confirmation before changing data.
My role: end-to-end design and development: workflow, tools, prompts, agent logic, integrations and production rollout.
NomadOS: the app for digital nomads
What it is: an app that brings together trip planning, visa tracking (with a Schengen 90/180 calculator), expenses in 150+ currencies, documents and a real-time community.
AI: visa recommendations, city guides and day-by-day itineraries generated with LLMs (Gemini, OpenAI).
My role: I designed and built it on my own, end to end: product, frontend, backend, integrations and store publication.
Stack: React + TypeScript, Supabase (PostgreSQL, Edge Functions, Realtime), Capacitor, Stripe and RevenueCat.
OCR + LLM invoice extraction
Problem: the firm manually typed all its clients' invoices into its accounting software.
Solution: a custom platform with a workspace per client and a portal where clients upload their own invoices. OCR reads each document, an LLM extracts the accounting data (invoice type, taxable base, VAT, withholdings, totals and more), which is validated and exported ready to import.
Result: hours of manual data entry eliminated every month, with human review only for doubtful cases.
LLM-based message classification and routing
Problem: client messages arrived in a single channel and had to be read one by one to decide which area they belonged to.
Solution: an LLM reads each message, classifies it by legal area and urgency, summarizes it, creates a task assigned to the area's lawyer, notifies them and confirms to the client that their case is with the right team. Ambiguous cases go to human review.
Result: manual first-level screening is gone and every enquiry reaches the right person sooner.
Multichannel data platform
Problem: sales and operations data was spread across Shopify, Amazon and Odoo, with no unified view.
Solution: a production platform that connects daily to the three APIs, loads the data into structured tables and powers SQL queries and dashboards.
Result: a single source of truth for sales and operations, updated automatically every day.
Financial reporting automation
Problem: daily reporting (P&L, exposure, liabilities, currencies) meant opening 30+ sources and extracting data by hand every day.
Solution: automated extraction with Python, APIs, RPA (Selenium) and n8n, with validations, reconciliations and anomaly alerts before reports are distributed.
Judgment: critical and regulatory outputs keep human review; everything else is fully automated.
Multimodal conversational product search
What it is: a search engine that understands natural-language queries such as "white dress for a communion, girl aged 8" and returns relevant products.
How: hybrid search (embeddings + BM25 with RRF), cross-encoder re-ranking and image embeddings, evaluated with Recall@k, MRR and NDCG against the current search engine.
My role: I lead the project (work plan, repository, team reviews) and build the LLM agent that interprets queries, plus the FastAPI API and Streamlit frontend.
How I can help your business
Tailored projects on top of the tools you already use. I always start with the business problem, not the technology.
AI agents & assistants
LLM agents that understand natural language and act on your systems.
Process automation
Workflows that remove repetitive work and connect your tools via APIs, with error handling and traceability.
AI document processing
Data extracted from invoices, contracts or forms with OCR and LLMs.
Custom software
Web and mobile platforms designed around the way you work.
Data & reporting
Pipelines that consolidate your data and generate reliable reports and dashboards.
AI consulting
I help you decide what to automate, what needs human review and what isn't worth it.
I understand the process and where time is lost.
I propose the solution and where it needs human review.
I develop, test with real data and integrate.
I deploy, measure the impact and iterate.
Madrid · remoteFrom finance to applied AI
I like finding the repetitive work nobody wants to do and turning it into something that runs on its own.
I started in data operations at a regulated FX brokerage, automating financial reporting with VBA, Python and RPA. Today I build LLM agents, integrations and custom software, and I created NomadOS, my own app.
My background is in Economics and International Relations: before building anything, I ask what business problem it solves.
Let's talk
Tell me which process you'd like to automate or what project you have in mind. I reply within 24-48h.