At Mastercard, I build applied AI capabilities for a transaction research platform, including RAG workflows over payment procedures and transaction references. I tuned retrieval and evidence selection to improve the context returned for analyst questions.
I built LangGraph workflows for transaction lookup, document retrieval, validation, and evidence return, and developed a React and TypeScript interface for reviewing citation-backed results. Routing straightforward lookups to a lower-cost model and reusing retrieval results reduced average inference cost by approximately 17% while maintaining the existing answer-quality threshold.
At Highmark Health, I built complaint-intelligence pipelines that classified member narratives and extracted entities for servicing teams. After reviewing false positives and missed escalations with stakeholders, I adjusted labels and thresholds, reducing unnecessary manual reviews by approximately 28% while maintaining approximately 95% recall for escalation-worthy cases.
Earlier, I worked on equipment demand forecasting and inventory planning at Charter Communications / Spectrum, flight disruption analytics at Delta Air Lines, and e-commerce analytics and machine-learning support at Amazon. Across these roles, I applied predictive modeling, forecasting, NLP, and production engineering to operational and business problems.
