At Adidas, I deployed Databricks real-time serving endpoints for Homefeed Personalization, achieving sub-100ms latency across different markets and increasing click-through rate by 27%. I also developed a product-attribute tagging pipeline that replaced manual tagging and a third-party vendor, decreasing inference costs by 50%.
My work has also included an XGBoost employee resignation model at ScotiaGBS, NLP search and computer vision applications at Equinox AI, and backend APIs at Asesoftware. At Universidad de los Andes, I supported Machine Learning courses and developed an interactive application for modeling and visualizing solar and biomass energy generation.
