Machine Learning Engineer | Data Scientist

• Built a reusable Infrastructure-As-Code platform using AWS Cloud Development Kit (Python), enabling the Vehicle Usage team to self-provision AWS resources, reducing dependency by 50-60% on the central MLOps team.

• Automated provisioning of data infrastructure, including Amazon Athena databases/tables, Amazon S3 storage, & AWS Glue catalogs, with IAM policies & CI/CD deployments enabling secure access for Amazon SageMaker workloads.

• Led the design & productionization of an end-to-end ML system detecting refueling events across 70–75M journeys per month, with scalable training & inference pipelines, feature engineering, data validation, & automated monthly retraining.

• Developed a K-Means pipeline, deployed via containerized services using Docker, Amazon ECR, AWS Fargate & EventBridge, with historical backfilling & BigQuery outputs delivering fleet analytics to cross-functional teams.

• Designed data-quality pipelines using AWS Glue & Athena with 4+ quality metrics & 5 QuickSight dashboards, trained an ARIMA anomaly detection model in SageMaker across 12B+ records for 70K vehicles, scalable to 7M+ globally.

• Optimized operational efficiency by eliminating external service costs to WirelessCar, enhancing VCDP data consistency, & reducing data quality-checking time by 20% for cross-functional teams.

• Developed a lane detection system using CNNs & Hough Transform, achieving 92% accuracy on 50 K+ images across diverse road & lighting conditions, ensuring robust performance in urban, highway, & low-contrast driving scenarios.

• Applied advanced image preprocessing techniques, including edge detection & perspective transforms, improving low-visibility detection by 25% & optimizing the pipeline for real-time inference & scalable deployment.