Led a multi-disciplinary content moderation team and scaled ML microservices for real-time content moderation with human-in-the-loop architecture, handling millions of requests per day. Developed transformer-based text classification model to improve safety and reduce offline transactions through user chats by 22%. Built and productionalized image search tool with React for the in-house admin tool, increasing true positive detection by 12%. Led systems to prevent keyword stuffing from inception to production, increasing orders per million impressions by 3% and reducing keyword stuffing reports by 75%. Developed ads ranking system with CTR models, increasing ad revenue by 4%, with Kubeflow and CI/CD pipelines. Led product recognition system to automate listing process. Architected and maintained a full-stack moderation system on Kubeflow Pipelines, increasing detection by 35% and reducing costs by 40%. Developed and A/B tested search ranking models with reproducible experiments and onli

Senior Software Engineer at Mapbox

January 1, 2024 - Present

Designed and deployed CI/CD pipelines to automate ML workflow—training, testing, and deployment with integrated reporting and alerting. Reduced deployment time by over 30% and improved model delivery reliability for real-time inference across four production services. Led development of a semantic query understanding module to identify brand, category, and place signals in queries, increasing CTR by over 7%. Collaborated with data teams to set up experiment tracking, model registry integration, and drift detection workflows.