Urjith Reddy Pudukaram

About Urjith Reddy Pudukaram

I\'m Data Scientist with 4+ years of experience building and deploying large-scale machine learning systems that power high-growth technology platforms. Currently working at Uber, where I develop advanced ML models that optimize marketplace efficiency, improve driver earnings, and enhance rider experience across millions of trips.My expertise lies in machine learning, reinforcement learning, and deep learning, with hands-on experience building models for dynamic pricing optimization, fraud detection, credit risk modeling, recommendation systems, and churn prediction. I specialize in transforming complex business challenges into scalable, data-driven solutions that deliver measurable business impact.At Uber, I developed reinforcement learning and gradient boosting models to optimize surge pricing, improving driver earnings by 12% and reducing rider wait times. I also built LSTM-based ETA prediction models using real-time traffic and trip data to improve prediction accuracy and marketplace reliability. Additionally, I design and evaluate experiments using A/B testing and causal inference techniques to guide product and marketplace decisions impacting millions of users.Previously at Pine Labs, I built real-time fraud detection systems and credit risk models that processed hundreds of thousands of daily transactions, achieving high accuracy and generating multi-million-dollar savings. I also developed recommendation systems, customer segmentation models, and time-series forecasting solutions to improve merchant retention and transaction success rates.Beyond modeling, I have strong experience in data engineering and MLOps, building scalable data pipelines using Spark, Airflow, and Kafka, and deploying models using Docker, Kubernetes, MLflow, and AWS with automated monitoring and drift detection.

Experience

1. Data Scientist

   Uber

   Jan 2025 — Present · CA, US

   Developed dynamic pricing optimization models using Reinforcement Learning and Gradient Boosting (XGBoost) to improve surge pricing accuracy, increasing projected driver earnings by 12% and reducing rider wait times by 8% in the Bay Area marketplace.• Built ETA prediction models leveraging real-time traffic, weather data, and historical trips using LSTM neural networks and XGBoost, reducing prediction error by 15% and improving rider satisfaction metrics.• Designed driver churn prediction models using behavioral and earnings data to identify at-risk drivers, enabling targeted retention strategies for top-performing drivers and improving marketplace stability.• Conducted A/B testing and statistical analysis for in-app product features impacting 2M+ weekly active users, delivering data-driven insights that informed product roadmap and marketplace optimization decisions.• Built causal inference models using propensity score matching and difference-in-differences to evaluate driver incentive programs, guiding allocation of $2M+ monthly incentive budgets.• Engineered scalable feature pipelines using PySpark and Databricks to process 50M+ daily trip records and deployed models via Uber Michelangelo (ML platform) with automated monitoring, drift detection, and real-time performance tracking.

Education

• Western Michigan University

  Master's degree, Data Science

• Amrita Vishwa Vidyapeetham

  Bachelor of Technology - BTech, Electrical, Electronics and Communications Engineering

  2019 — 2023