Data Scientist · Applied AI
Valencia, Spain · available for remote work
I turn business processes into data tools my team uses every day: rules on top of the ERP, dashboards with alerts and search that gives direct answers.
I'm finishing an MSc in Data Science (UOC, Feb 2027) with a computer vision thesis.
Tools in use at my company and personal projects. The demos use synthetic data.
A Python system that collects AI news, filters it, checks every claim against its source, writes, edits and prepares each edition. A deterministic orchestrator I built decides the order; if an AI engine fails, another takes over. Publishing is always a human decision.
One every morning, with 392 automated tests.
A gym, cardio and nutrition app that logs every set, detects personal records live and suggests the load for the next session. It works without signal and syncs later. 14 people use it.
Room is the source of truth and Firestore the mirror. A quota audit cut each sync pass from about 2,800 reads to 52.
A single text box for customer service: price and stock per variant, the right spare part, whether a shower tray fits a shower screen, or what to do when “the tap is leaking”.
Inverted index, edit distance and BM25 with trade synonyms. Deterministic, no cost per query and works offline.
Type like a customer would, typos included. The search shows what it understood and answers with price, stock, the shower screen that fits an opening or the spare part.
Same engine, interpreter and fictitious catalogue as the full demo. Try it in Spanish: the catalogue and the queries are in Spanish, so each example shows what it means.
Loading the fictitious catalogue…
A rule engine reads the ERP export twice a day and gives customer service a work list: what to review, why it flagged and who has already validated it.
Five separate reports became one dashboard that updates itself from SAP, compares each carrier and each incident reason with its own history, and greys out anything with little data.
Two extensions I built for Whisper Money, an open-source finance app: a simulator that compares where an early repayment saves the most, and an MCP server that only lets the assistant run 19 specific operations and asks for confirmation before any change.
A dashboard that brings together finances, nutrition, training and my MSc, calculates weekly indicators and suggests priorities with explicit rules.
It gathers job offers from several public sources, merges them and scores them with six factors and five checks. It sends nothing: it prepares a report and leaves the decision to a person.
Thesis in progress: turtle species identification with computer vision. I am building a pipeline that will automate species identification and the location of anatomical landmarks that are currently marked by hand, on a real dataset of about 150 turtles.
