Article · 2026
Injecting company jargon for generative AI
How to inject company-specific knowledge (teams, project phases, glossaries) into LLM prompts so they turn natural-language requests into valid arXiv API queries, tested across model sizes.
Senior Data Science Consultant · SDG Group
Astrophysicist turned data scientist. I build GenAI, NLP and computer-vision solutions, and explain them to people who don't live in notebooks.
Madrid, Spain
I spent six years analysing X-ray and optical spectra of supermassive black holes before moving into industry. Today I’m a senior data science consultant at SDG Group, where I develop Data Science and AI solutions for companies and take on technical guidance and AI lead tasks. Among the projects I’ve been involved in, I’ve built RAG chatbots used in production, generative-AI assistants for logistics planning, computer-vision defect detection and statistical studies for clients in banking, life sciences, insurance, food and energy.
That scientific background still shapes how I work: I care about getting the question right, measuring things honestly and communicating results succinctly to non-technical audiences. I’m a curious, fast learner and I enjoy mentoring juniors (I’ve mentored graduates in data science and computer vision and directed a Bachelor’s thesis).
From X-ray spectra of black holes to generative AI. Each absorption line is a milestone: hover over or tap the spectrum to explore them.
Sep 2024 – Present
SDG Group · Madrid, Spain
GenAI solutions for a pharmaceutical company, from a research-assistant chatbot to grounded content generation, raising user satisfaction by around 40% over the previous tools. Technical guidance and team lead tasks.
May 2021 – Aug 2024
Cognizant Technology Solutions · Madrid, Spain
Generative AI, NLP, computer vision and statistical projects for clients in life sciences, banking, insurance, food quality control and oil & gas.
Jan 2014 – Dec 2019
Institute of Physics of Cantabria (CSIC-UC) · Santander, Spain
PhD research on the obscuration of Active Galactic Nuclei using X-ray and optical spectra from the Bright Ultra-hard XMM-Newton Survey.
Article · 2026
How to inject company-specific knowledge (teams, project phases, glossaries) into LLM prompts so they turn natural-language requests into valid arXiv API queries, tested across model sizes.
Talk · 2024
Invited speaker at the International Conference on FOSS Approaches towards Computational Intelligence and Language Technology (ICFOSS, India, March 2024), presenting how LLMs perform on mathematical operations.
Case study
A general fruit-detection model on Microsoft's Azure Percept edge-AI device, using Custom Vision and IoT Hub. Published on Microsoft's Internet of Things blog.
Case study
Deep-learning models for automatic surface-rust detection in industrial assets, part of a Cognizant solution aiming to cut inspection costs by 50%.
Before data science I was an astrophysicist. My PhD at the Institute of Physics of Cantabria studied how dust and gas hide the light of Active Galactic Nuclei (supermassive black holes growing at the centres of galaxies), combining X-ray and optical spectra of hundreds of objects.
My first machine-learning work was also in astronomy: a faster, single-pass k-means for classifying large sets of galaxy spectra.
When I'm not working with data, I paint miniatures. Here are a few of the good ones
Happy to talk about data science, generative AI or black holes.