At Modo, I own the end-to-end lifecycle of a personalized promotion ranker, from feature engineering and LightGBM LambdaRank training to deployment on AWS SageMaker. Three iterations delivered a statistically significant 4.6% CTR lift and 1.1% CVR lift, validated through A/B testing.
I’ve also built natural-language promotions search with hybrid retrieval and LLM-based query understanding, and created evaluation and MLOps foundations for production systems. Previously, I led data science work at Blend360 for Wyndham and at Metadata.io, including an LTV model that improved forecast accuracy by 12 R² points and churn analytics that supported a 12% increase in average client spend in one quarter.

