Gaddam Siva Kumar

Multilingual Comment Classification Using TPUs to identify toxicity in comments. Deployed classification model in GCP pre-production using Docker image. Used GKE for container management, load balancing, and recovery. Automated testing and deployment with Cloud Build. Created Looker Dashboard for model predictions. Implemented data drift detection and retraining using ML Flow and Cloud Functions. Developed and deployed demand forecasting models for PayPal support agents. Built data drift detection systems and data marts using Spark and Airflow for online agreements notifications. Conducted path analysis for customer journeys and developed dashboards for key metrics.

Developed OCR Driver ML Model for automating driver onboarding using ABN document validation. Improved OCR accuracy from 80% to 96% through image processing and deployed model to production. Implemented real-time driver fraud detection system using machine learning models, Docker, and Kubernetes. Built automations for API integrations and web scraping. Developed forecast models for inbound tickets and capacity planning. Deployed market share forecast models for ANZ business.

Jewellery SKU Forecasting for Pandora Jewellery using traditional and advanced ML methods, improving accuracy from 78% to 88%. Developed look-a-like modeling and clustering for new SKUs. Conducted funnel analysis to improve payment page conversion rate. Implemented rider growth optimization by reducing cancellation rates and increasing driver supply hours.

Built dashboards for Mexico End-User Insights with hourly refresh and drill-down capabilities. Assessed campaign metrics and performed exploratory data analysis. Developed SQL procedures and triggers for insurance underwriting data ingestion and movement.