At REPSOL, I benchmark Classic RAG, GraphRAG, LightRAG and agentic retrieval approaches, assessing retrieval quality, response accuracy and execution performance. I also built a local Model Context Protocol server integrated with Neo4j.
At AFI, I estimated mortality probabilities with logistic regression and generalised linear models, and developed stochastic and numerical models for financial option valuation. I also built a personal multi-agent Python pipeline with local LLMs to generate validated reports.
At ACCENTURE, I supported client data initiatives and contributed to analysis and reporting, presenting results to teams weekly. My work there also included collaborating in daily project meetings.
My studies at CUNEF Universidad combine Mathematical Engineering and Economics. In an independent fraud-detection project, I compared XGBoost, CatBoost and TabPFN pipelines, evaluating class imbalance, decision thresholds and SHAP explanations.

