Background#
I’m Xoel García Maestu, a 4th-year student in the BSc in Data Science and Engineering at the University of A Coruña (UDC).
I work with the full data lifecycle: from capture and preparation to modeling, deployment, and visualization. I combine Data Engineering and Statistical Analysis with Artificial Intelligence and Deep Learning techniques (computer vision, NLP, federated/continuous/distributed training…) applied to real-world use cases.
Before university, I participated in science projects and fairs during high school, which accustomed me early to teamwork, preparing presentations, and defending ideas publicly. Now I apply that same approach to data and AI projects in academic and professional settings.
Work Experience#
NTT DATA · Data Analytics & BI Intern#
Sep 2025 – Nov 2025 · A Coruña · Hybrid
Curriculum internship in technology consulting for the banking sector, within a workforce data analytics project.
Main responsibilities:
- Development of Python data pipelines for cleaning, integration, and validation.
- Construction of historical datasets and generation of synthetic data for testing and analysis.
- Design of Looker Studio dashboards tailored to business users.
- Automation of periodic information delivery through scripts.
- Migration of Google Apps Script (JavaScript) logic to Python for improved maintainability and traceability.
This period helped me understand banking business processes, see closely how its internal structure is configured, and how technical decisions relate to business requirements and real data limitations.
Awards & Recognition#
1st Prize – Datathon “O Camiño dos Datos” (2026)#
On January 30, 2026, we presented MXP AutoScan at the final day of the “O Camiño dos Datos” Datathon, organized by DIHGIGAL, ITG, and the University of A Coruña, funded by IGAPE. We won first prize before a jury specialized in technology-industrial fields who valued the relevance of the challenge addressed, data quality usage, obtained conclusions, and real impact of the proposed solutions.
Our project, MXP AutoScan, is a computer vision solution using Dual Mask R-CNN to automate vehicle assessment. The system analyzes images to detect and locate damage, identify affected parts, and estimate severity, generating technical reports through a web interface or REST API designed for insurance companies, workshops, and mobility businesses to streamline and make transparent the evaluation process.
Read more about our project in El Español’s coverage.
CanSat Program (ESA / ESERO Spain)#
In 2021 I was part of the BujanSat team in the European Space Agency CanSat project. The challenge consisted of 2 missions:
- The main mission was the same for all teams: replicate a satellite’s main subsystems within the reduced space of a soda can. This includes taking real-time measurements of temperature, humidity, atmospheric pressure, and acceleration during launch with a sounding balloon or small rocket (300-800m altitude). The system had to autonomously log data, transmit live telemetry via radio, and deploy a parachute for controlled descent.
- The secondary mission was free choice for each team. We built a seed dispenser that activated during CanSat descent as environmental awareness for wildfires that Galicia typically suffers during heat seasons.
After winning 1st place in both the Galician regional phase and Spanish national phase, we represented Spain at the European phase. We won the Best Outreach award for project communication and outreach.
Science Fair Awards (High School)#
During high school I participated in science, technology, and outreach projects: microplastic measurement on Spanish beaches, DIY 3D-printed biotechnology lab, CanSat, or Snack Experiments for science outreach.
With these projects we achieved awards at national and international science fairs, including first prizes or honorable mentions at Festiciencia, Diverciencia, Ciencia en Acción, Galiciencia, Maker Faire, OpenScience Cambre, and Vigo’s Intercommunity Youth Research Forum.
Areas of Interest & Technologies#
From the beginning of my Data Science and Engineering training, I was drawn to understanding data in its real context: how it’s generated, structured, and turned into useful decisions. I built a foundation in three complementary areas: data engineering that supports data systems, statistics and analysis that extract value from them, and machine learning that automates and scales that value.
I primarily work with Python, complemented by R when needed for statistical tasks, Java for information retrieval systems, Julia for numerical computing, and JavaScript for web visualization and integration.
Throughout my degree I’ve worked with technologies and concepts from these areas:
Data Engineering: Design and implementation of ETL pipelines, modeling of operational and analytical databases, and distributed processing with Apache Spark and Spark MLlib. Used SQL, analytical SQL, PostgreSQL, MySQL, Oracle, MongoDB, Snowflake, and PostGIS for spatiotemporal information, with heterogeneous source integration via Denodo. Git and Docker as standard tools, and AWS for cloud computing.Data Science: Statistical knowledge in inference, hypothesis testing, linear/polynomial/logistic regression and nonparametric regression, high-dimensional data modeling, time series (ARIMA), and simulation/resampling techniques (bootstrap). Exploratory analysis with NumPy, Pandas, Polars, PySpark, Matplotlib, Seaborn, etc. Results communication through dashboards in Power BI, Looker Studio, and Tableau, with MDX query experience on OLAP cubes.Artificial Intelligence & Deep Learning: Classical machine learning with Scikit-learn and Imbalanced-learn, bagging models (Random Forest) and sequential boosting (AdaBoost, Gradient Boosting, XGBoost, LightGBM) to optimize bias-variance tradeoff alongside hyperparameter optimization libraries such as Optuna or Hyperopt. Deep models with PyTorch, Keras, and TensorFlow: vision architectures (CNNs, ResNets, DenseNets, U-Nets, Mask R-CNNs), temporal sequences (LSTMs, GRUs), and generative models (Autoencoders, VAEs). Experience with transfer learning and fine-tuning on pretrained vision (MobileNet, ImageNet) and language (BERT) models via HuggingFace. Model lifecycle tracking with MLflow. Also worked with paradigms like semi-supervised learning, reinforcement learning (Q-Learning, SARSA), federated, continuous, and distributed learning, as well as NLP, and information retrieval with Apache Lucene.
Beyond the Screens#
Outside strictly academic areas:
- I’m interested in outdoor sports like surfing or climbing. I’ve competed in sport climbing (bouldering, lead, speed), with Galician podiums and Spanish championship finals. This has helped me work with discipline toward goals while being aware physiologically, cognitively, and emotionally.
- I’ve played clarinet for over 10 years. I started at Escuela de Música Sementeira and continued at A Coruña Professional Music Conservatory, completing both Elementary and Professional Degrees. I have experience in bands, chamber music, and solo/accompanied performance.
- I’ve volunteered at science fairs, presenting projects, supporting organization, managing social media, and conducting live interviews with scientists.
- I’m interested in documentary recording and art in its various expressions: audiovisual, pictorial, sculptural, installations, performances, relational, etc.
Contact#
- Email: xoel.maestu@gmail.com
- LinkedIn: linkedin.com/in/xoel-maestu
- GitHub: github.com/xoel-maestu