Real-time recommendation engine development with end-to-end ML Ops pipeline using TensorFlow and automated deployment. Implemented secure access controls and managed deployments on AWS, ensuring enterprise data governance. Monitored model drift with MLflow and established continuous evaluation and retraining workflows to improve prediction reliability. Targeted sub-15ms inference latency in production environments and supported AI delivery best practices across teams.

ML Associate at PwC

July 1, 2024 - April 1, 2025

Collaborated with data scientists and business stakeholders to translate analytical requirements into production-ready machine learning solutions, leveraging MLflow to track experiments and manage pipelines. Designed and evaluated ML models using Scikit-learn for customer retention and demand forecasting, resulting in improved prediction accuracy by 18% over prior approaches. Implemented feature selection, hyperparameter tuning, and model validation experiments to improve model performance, and communicated findings and recommendations to business stakeholders and project leadership. Drove deployment and monitoring of ML models in production, tracking performance trends and helping maintain model accuracy within 5% of validation benchmarks after release. Built end-to-end ML workflows and automated parts of the data pipeline to accelerate delivery.

Data Engineer / Scientist at Cognizant

September 1, 2019 - May 1, 2023

Built Spark-based ingestion pipelines processing customer and transaction data from multiple source systems, reducing data preparation time for reporting teams by 40% and improving downstream data availability across enterprise reporting platforms. Collaborated with business analysts and product stakeholders to translate reporting requirements into scalable SQL-based data models, supporting over 100 recurring dashboards used by finance and operations teams. Owned production ETL workflows, troubleshooting pipeline failures and optimizing transformation logic to improve job success rate from 92% to 99% during peak processing periods. Partnered with senior data scientists to prepare training datasets, engineer predictive features, and automate data quality checks, reducing model development time by nearly 30%. Developed and evaluated ML models using Scikit-learn for customer retention and demand forecasting initiatives, improving prediction accuracy by 18% over existing rule-based approac