Professional Experience
Automation Consultant
- Design and development of digital systems and process automations for companies on Airtable, Notion, Monday and Slack, using n8n and Make, owning the whole project: system design, technical development and client communication.
- Audit of clients' internal processes with improvement proposals, and AI training for companies.
Predoctoral Researcher
- Building and evaluation of classification models for massive stars in the infrared (PCA, XGBoost, Random Forest) from He and H diagnostic lines, in Python, end to end: data, model, evaluation and documentation.
- Statistical analysis of results and validation of conclusions against observations and the literature.
- Presentation of results at international conferences in English.
Engineering Intern, Structures department
Education
PhD in Astrophysics
MSc in Astrophysics
BSc in Aerospace Engineering
Technical Skills
Programming
Python, R, Matlab, SQL, Git, Docker.
Statistics and machine learning
Statistical inference, hypothesis testing, PCA, classification models (XGBoost, Random Forest; scikit-learn), model evaluation and validation, time series, Monte Carlo simulation.
AI and automation
Local LLMs (Ollama, Qwen), Anthropic API, MCP (Model Context Protocol), n8n, Make, Airtable, Notion, Monday, Slack.
Office
Excel, Word, PowerPoint.
Languages
Spanish (native) · English C1 (Trinity College London) · French A2.
Certifications and Courses
CFA Level I
Model Context Protocol: Advanced Topics
n8n: Zero to Hero
IBM Data Science Professional Certificate
Deep Learning Specialization
Publications
Peer-reviewed papers. Full list on ADS (ORCID 0000-0002-6336-0359).
First author
First spectroscopic identification of the main sequence in Westerlund 1
Astronomy & Astrophysics, 708, A209 · DOI 10.1051/0004-6361/202558099
Castellanos et al. 2026b
Astronomy & Astrophysics, accepted (link to follow on publication)
Diffuse interstellar bands in the near-infrared: expanding the reddening range
Monthly Notices of the Royal Astronomical Society, 532, 2065–2081 · DOI 10.1093/mnras/stae1472
Co-author
Projects
Personal projects in production (2026). All the code is at github.com/CasteSan.
Job-offer radar
A daily radar that collects job offers, filters and scores them with a local language model, and sends a prioritised digest. The scoring is versioned and evaluated against human verdicts, and the system is tested against prompt injection.
PortfolioHub — investment portfolio tracker
A web application to track investment portfolios: Spanish tax engine (FIFO/IRPF), TWR/MWR returns, risk metrics, ECB exchange rates and broker import parsers. It runs in production on Vercel; the pure calculation engines are public, with more than 500 tests and screenshots of the dashboard.
E-mail classifier with a local LLM
An e-mail classifier that runs a local language model inside an n8n workflow on Docker and reports through Telegram.