Mahesh Kumar Duvvuri

About

Backend/infra engineer with experience across Nestle, J.P. Morgan, and Capgemini, combining ML systems work (YOLOv8/PyTorch object detection with TFLite edge deployment) with production-grade cloud/Kubernetes operations. Has delivered measurable impact via AWS migrations (25% cost reduction, 99.9% availability), microservice modernization (35% faster processing), and low-latency Kafka streaming for financial dashboards (<100ms) using DLQs and idempotent consumers.

Experience

Education

Key Strengths

• Architected end-to-end object detection pipeline for edge, including TFLite conversion for low-latency inference
• Designed modular ML system separating training, evaluation, and inference for independent scaling
• Performance optimization: hyperparameter/batch tuning to maximize GPU utilization and improve mAP
• Deployed Dockerized microservices on Kubernetes (AWS EC2) with health checks and autoscaling for peak traffic reliability
• Built CI/CD with Jenkins + GitHub; automated provisioning with Terraform (50% setup time reduction) and integrated SonarQube security scans
• Implemented centralized logging/observability with ELK to reduce debugging time
• Led on-prem to AWS migration of legacy apps (25% hosting cost reduction; 99.9% availability)
• Modernized legacy reporting into Dockerized microservices (35% faster data processing)
• Real-time Kafka streaming for financial dashboards with sub-100ms latency requirement
• Resilient streaming design using DLQs and idempotent consumers to ensure financial data consistency
• Strong GitOps practice: Git as source of truth, PR-based infra changes, automated applies to prevent configuration drift