Norrköping, Sweden · Open to data, AI and automation roles

Data analyst and automation developer.

Two kinds of work: making data good enough to decide on, and taking the manual work out of a business. A BSc in Sociology, an MSc in Computer Social Science, four years of ETL and analytics in production.

See the work Download CV (PDF)
Mario Soriano Bañuls, data analyst and automation developer
27validated tables from INE and Eurostat, one pipeline
3.8 → 1.8Spain's inequality gap with the EU, 2015 to 2024
−6.97%real income at the bottom decile since 2008
2client systems live in production

01 · What I do

Three things, over and over

Most jobs are one of these three wearing a different hat.

01

Data that does not add up

CRM, ERP and till data come from three systems that speak different languages. I build the pipelines that join them, validate them, and surface the few numbers the business runs on.

02

Public data turned into a decision

Inequality, poverty and mental health are all published as raw tables. I extract them, check them against each other, and test whether the story people tell is the one the data supports.

03

Systems that run themselves

Order capture, margin control, document intake. Built once, then they run overnight without someone watching, behind tests and a deploy pipeline.

02 · Selected work

Five projects, five different problems

Three are mine. Two run inside a client business and carry their name. Every number below is read off the data or the screenshot next to it.

Line chart comparing the Gini coefficient in Spain against the EU27 average from 2015 to 2024. Spain falls from 34.6 to 31.2 while the EU average falls from 30.8 to 29.4, so the gap narrows from 3.8 to 1.8 points.

The gap with the EU has more than halved, and still sits at 1.8 points.

Bar chart of the Gini coefficient by Spanish region in 2023. Melilla is highest at 41.1 and Aragon lowest at 27.2, a spread of 14 points, against a national figure of 31.5.

A 14 point spread inside a single country.

Project 01 · Inequality in Spain

Can you prove inequality is actually improving?

INE and Eurostat, 2002–2025 · SQL Server, Python, GitHub Actions · mine

The problem. Everyone has an opinion about whether Spain is getting more or less equal, and nobody could show which. The numbers existed; nobody had joined them up.

What I built. 27 tables from 13 INE series and Eurostat's EU-SILC into SQL Server, with declarative validation rules and a pipeline CI runs nightly. The dataset ships as CSV, so the conclusions can be checked without a database.

What it showed. The gap with the EU fell from 3.8 to 1.8 points between 2015 and 2024. But the bottom decil lost 6.97% of real income since 2008, against 4.99% at the top. Converging with Europe, worsening inside the country.

27tables, 13 INE series plus 14 Eurostat
+1.8 ppSpain versus the EU, 2024
−6.97%real income, bottom decile, since 2008
14 ptspread between the most and least equal region

See the code and the data dictionary

Interactive map of Spain shaded by deaths from mental disorders per 100,000 people by autonomous community in 2022, with a time series from 1980 and filters for sex and cause of death.

Published on Tableau Public. Built in Spanish, so the interface is in Spanish.

Project 02 · Mental health

A national average that hides the map

Source: INE · Tableau · mine

The problem. Spain's mental-health trend gets quoted as one national figure, which averages away the regions where the difference is.

What I built. An interactive choropleth of deaths from mental disorders per 100,000 by autonomous community, with the series back to 1980 and filters for sex and cause.

What it showed. Deaths per 100,000 went from 2,155 in 1980 to 5,040 in 2022, the worst on record. Between regions the range runs 1.18 to 15.52.

5,040deaths per 100,000 in 2022
13×difference between the best and worst region
42 yrsof series, 1980 to 2022
INEofficial source, national statistics institute

Open it on Tableau Public

Power BI dashboard breaking down a survey of 468 data professionals by gender, salary, education and country, with pay satisfaction rated 4.27 out of 10 and work life balance 5.77 out of 10.

Pay by role, satisfaction and balance across 468 respondents.

Project 03 · Work in data

What the work is actually like

468 respondents · Power BI · mine

The problem. Whether data work is worth doing gets answered with anecdotes. Most of them are second-hand.

What I built. A Power BI breakdown of salary, satisfaction and education across 468 people in the field, sliced by role, gender and country.

What it showed. Pay ranges roughly twofold across roles, and nobody is satisfied: pay 4.27 out of 10, work-life balance 5.77.

468people surveyed in the field
2×pay spread between top and bottom roles
4.27satisfaction with pay, out of 10
5.77work-life balance, out of 10
Closing screen for a trading day showing 42.40 kg sold, 1073.15 euro cost, 1323.77 euro revenue and 250.62 euro margin across 19 product lines.

Real client interface. The product names are the client's own data, so it stays in Spanish.

Traceability table linking each species sold back to the supplier delivery note that paid for it, with a match confidence of at least 95 percent.

Every species line links back to the note that paid for it.

Project 04 · Francis Pascual

Knowing the margin on the day it happens

Fishing cofradía, Alicante · live · built and maintained solo

The problem. Fish is bought by weight, sold by species and invoiced by supplier. Prices move daily, and the margin was worked out by hand, weeks after the fact.

What I built. Emailed delivery notes are read into structured records and checked against that supplier's own history, which catches duplicates and impossible amounts. From the verified cost a price is derived, and anything under target is flagged.

What it changed. The cofradía knows its margin the same day, per species and per supplier, instead of at the end of the month.

42.40 kgsold in a single trading day
1,323.77 €revenue, from a 1,073.15 € cost
250.62 €real margin across 19 species
0lines sold below the target price
A conversation with an AI shopping assistant for an online fruit retailer. The customer asks for potatoes and oranges and the assistant returns an order summary with catalogue prices, a shipping cost, a total of 19.09 euro, and asks for confirmation.

A real conversation from the live assistant. The Spanish is the product working, not an untranslated screen.

Project 05 · Cosechados

An assistant that takes the order while you sleep

Online fruit retailer, Spain · live on the client's own site

The problem. A fruit retailer sells at night and at weekends, when nobody is on the invoice, and the catalogue lives in one person's head.

What I built. An assistant that answers from the client's own catalogue and documents, and returns its sources. It takes a whole order in ordinary conversation, prices it, applies the shipping threshold, and waits for confirmation.

What it changed. Nothing is lost overnight, and every price quoted can be traced back to where it came from.

45 €shipping threshold applied unaided
2channels on one brain: web and Telegram
1 doconboards the next client
liverunning on the client's own domain

Try the live assistant

03 · Background

Where the work happened

2026 – present
Babieca Tech
Co-Founder · Automation and data

Client systems for small businesses: margin control, order capture and document intake. Two run in production. Everything ships behind a test suite and a visual regression pass.

2024 – 2025
RockStar Data
Data Analyst

ETL joining CRM, ERP and point-of-sale data, with automated validation that cut data-quality incidents. Reported the KPIs behind pricing decisions.

2026 – 2028
Linköping University
MSc Computer Social Science

Statistics and machine learning applied to social questions. Where the inequality work comes from.

2019 – 2023
University of Alicante
BSc Sociology

How to read a population, and how not to overclaim from it.

PythonSQLSQL Server pandasscipystatsmodels Power BITableauExcel

Data: SQL Server, PostgreSQL, Parquet, CSV. Analysis: SciPy, statsmodels, pandas, Jupyter. Visualisation: Tableau, Power BI, Matplotlib. Delivery: GitHub Actions, Docker, ETL, CI.

04 · Contact

Open to work in and around Norrköping

Native Spanish speaker, English at IELTS Academic C1. Data, AI, automation or software, remote or on site.

datamario.ba@gmail.com LinkedIn GitHub

CV · Pick the relevant one

Data, AI & Automation PDF Computer Social Science PDF International / part-time roles PDF