↳ DATA ENGINEER · BARCELONA
_
Turning raw data into scalable systems and real-world impact.
Profile
Credentials
Technical
Career
First professional experience in data analysis within a fast-paced insurtech startup. Designed and maintained interactive dashboards using Power BI and Looker Studio to support strategic decision-making, and automated manual reporting workflows using Python and Google Sheets. Built predictive models to optimize the expansion of Kanguro point networks, leveraging package volume and demand trends.
Promoted to Junior after demonstrating strong impact. Owned end-to-end KPI dashboards using Power BI and Looker Studio, and managed relational databases (SQL), ensuring data integrity, consistency, and performance. Contributed to database migration projects and built scalable ETL pipelines using Python to transform data from transactional systems into a centralized data warehouse. Applied data modeling best practices to support analytical use cases and collaborated closely with Product and Operations teams to enable data-driven decision-making.
CaixaBank
• Security Control & Access Management Migration:
The migration of CaixaBank's security control data environment was carried out from on-premise infrastructure to Google Cloud Platform (BigQuery). The migration strategy was defined, and end-to-end data workflows were automated using Apache Airflow, improving the reliability and scalability of data pipelines. The process was supported by SQL queries for data construction and transformation, and a Python layer was added to automate and orchestrate workflows from Cloud Composer.
• ETL Pipelines & Dashboarding Ecosystem:
Built and maintained ETL pipelines within on-premise environments, integrating multiple data sources (Qlik, R, Jupyter Notebooks) to support analytical workloads. Contributed to the development of interactive dashboards in Qlik and ensured the maintenance and reliability of existing reporting solutions used by business stakeholders for decision-making.
Artifacts
• Metadata Lineage:
Developed a metadata lineage model designed to extract and centralize metadata from multiple data sources, including BigQuery, MySQL, and PostgreSQL. Built a metadata artifact layer to structure and unify metadata information, enabling a data lineage system with propagation of metadata across tables and data models to improve traceability and governance. The development was implemented using Python.
Work
Intelligent system for Texas Hold'em that classifies poker hands and estimates win, draw, and loss probabilities using a CNN trained on a custom-generated dataset. Full-stack web app with a Flask backend and Angular frontend combining deep learning with Monte Carlo simulations.
End-to-end pipeline for detecting and recognising Spanish license plates through plate localisation with YOLOv5, classical image processing for character segmentation, and a hybrid CNN + SVM classifier achieving high accuracy across real-world lighting conditions and perspectives.
Real-time vehicle tracking and counting system integrating YOLOv5 and YOLOv8 with DeepSORT and Kalman filtering for consistent identity tracking across frames. Motion-based filtering and reference-line logic determine vehicle direction while frame skipping and image cropping achieve near-real-time performance.
Automated detection of H. pylori in whole slide images using two complementary approaches: an AutoEncoder-based anomaly detection model trained exclusively on negative samples, and an attention-based model for direct patient-level classification aggregating patch-level features.
Automated seizure detection from multichannel EEG data using CNN with spatial and channel attention, LSTM for temporal modelling, and a hybrid CNN-LSTM architecture achieving the strongest recall on the CHB-MIT clinical dataset under both population-based and personalised cross-validation strategies.
Spotify's related-artist network crawled using BFS and DFS algorithms to build directed, undirected, and weighted graphs across 384 artists. Girvan-Newman community detection achieved modularity 0.722, with centrality analysis revealing that network hubs and information brokers are structurally non-overlapping roles.
Graph database modelling historical census data (1833–1889) from two Catalan municipalities using Neo4j and Cypher. Includes 13 thematic queries covering kinship networks and cross-census identity resolution, plus GDS library analysis for Weakly Connected Components and PageRank centrality.
Document-oriented NoSQL database for a comic book publishing system, restructuring an ER model into two MongoDB collections using extended reference and attribute design patterns. Includes a Python ingestion script and ten aggregation pipeline queries covering cross-collection lookups and complex array filtering.
Comprehensive recommender system applying Item-Item collaborative filtering, SVD matrix factorisation, and Content-Based filtering on large-scale Amazon datasets across Books, Electronics, and Beauty. Models evaluated with Precision@K, Recall@K, MAE, RMSE, and NDCG to identify the most scalable approach.
End-to-end study comparing Moderna's global vaccine distribution against Pfizer, AstraZeneca, and Sinopharm. Data cleaning and EDA in R; interactive dashboards, animated maps, and time-series charts in Tableau and Shiny correlating vaccination rates, case counts, and socioeconomic indicators.
Full statistical data analysis in R covering data cleaning, exploratory analysis, and inferential statistics. Applies descriptive techniques, hypothesis testing, and visual exploration to extract insights from real datasets, with results documented in a structured reproducible report.
Get in touch
I'm open to new opportunities — full-time, hybrid or remote.
Feel free to reach out through any of the channels below.