At Traackr, I led development of Content Categorizer, an NLP classification system processing 10M+ posts per month across 50+ countries, with F1 above 90%. I also built an evaluation framework to measure AI agent quality and reliability.
I architected a production multi-agent system for text, audio, image, and video, and used LLMs to generate training data that improved model coverage and reduced manual annotation. I also worked with software engineering to optimize content pipelines, cutting costs by over 20%.
At Fintonic, I led a small data team and developed credit scoring and risk models under financial and GDPR requirements. Earlier, as a Data Science Consultant at Sintec, I delivered forecasting, inventory optimization, and pricing models for clients across LATAM.
