Developed and deployed deep learning models for Amazon’s recommendation and personalization systems, leveraging TensorFlow and SageMaker to drive an 18% increase in user engagement as measured by A/B testing. Built scalable ETL data pipelines using AWS services (Lambda, S3, Redshift, EMR) for demand forecasting and real-time analytics on >50M records/day, ensuring high system reliability (99.9% uptime). Reduced model inference latency by 30% (200ms → 140ms) via architectural optimization and TensorFlow Serving, improving the speed and relevance of product search and Browse Nodes. Led A/B testing, model monitoring, and implemented CI/CD (AWS CodePipeline, Docker) for rapid and safe deployment of ML models, cutting model release cycle time by 40%. Collaborated with engineers, DevOps, and product managers to integrate ML APIs into retail microservices, reducing support tickets related to recommendations by 20%.

Senior Machine Learning Engineer at IQVIA

May 1, 2021 - July 1, 2022

Designed predictive models for patient risk stratification and clinical outcome prediction in large-scale healthcare datasets, using Python, Scikit-learn, and TensorFlow to support population health and real-world evidence (RWE) studies. Automated feature engineering and data cleaning pipelines for multimodal clinical data (EHR, claims, genomics) using SQL and Pandas, improving model accuracy by 13% and speeding up time-to-insight for research teams. Led validation and documentation of ML models to ensure compliance with HIPAA, GxP, and IQVIA data privacy protocols, successfully passing all internal and external audits. Mentored a team of 5 engineers, conducting code reviews and training sessions to align development with industry best practices (MLOps, documentation, reproducibility).

Machine Learning Engineer at JPMorgan Chase & Co

April 1, 2018 - May 1, 2021

Developed and deployed machine learning models for fraud detection, credit risk scoring, and customer segmentation, using Python, Scikit-learn, and XGBoost to reduce false positives by 22% on real transaction data. Automated data preprocessing pipelines with Airflow and SQL, improving data quality and enabling continuous model retraining for evolving financial risk environments. Deployed ML models in production using Docker and enterprise CI/CD (Jenkins, Kubernetes), ensuring compliance with regulatory standards and internal risk controls. Presented technical findings and actionable insights to senior business and risk management stakeholders, directly impacting AI-driven strategies in digital banking, payments, and lending.