At PayPal, I architected and implemented a multi-agent, LLM-driven AutoML framework that handles dataset ingestion, preprocessing, training, evaluation, and explainability from natural-language requests. It reduced manual ML pipeline setup time by 70%.
I also developed a proxy-tag-based BNPL launch model for cold-start markets. Comparative experiments showed a 15 percentage-point AUC improvement over the legacy model for new applicants.
At Walmart, I architected DataBot, an LLM-powered platform that turns natural-language questions into SQL and delivers insights through a conversational interface. Its estimated value was $1 million per year through saved associate work time.
At Walmart, I also developed models for conversion and cancellation propensity, with experiments and outcomes tied to improved conversion precision, reduced churn, and incremental revenue. Earlier, at Latentview, I worked on NLP-based service-request classification and an interactive bot that reduced report-generation time by 70%.
