At Thomson Reuters, I designed and shipped an XGBoost model that categorized legal and tax publications for Checkpoint, replacing manual editorial triage. It earned first place in the Legal category of the Thomson Reuters Global AI Hackathon 2025.
I also built a Retrieval-Augmented Generation pipeline over the Checkpoint corpus, enabling natural-language queries with source-cited answers. I architected a multi-agent system on AWS Bedrock and TRACK to track and classify errors across Checkpoint microservices and suggest remediations.
At BBVA, I developed backend services in Java and Python for relational data management, integrated the LinkCard account service into the Youngsters app, and contributed to document safeguarding in SEIRI.
Earlier, at Junta de Andalucía, I led development of microservices and web services for government applications. I also built a FastAPI and scikit-learn service that classified incoming citizen requests and routed them to the appropriate department.
