Sarath Dunga

About

Full-stack engineer (~3 years) with eBay production experience building and operating high-scale, event-driven Python microservices for order processing and AI-powered recommendations (Kafka/Redis/FastAPI on AWS with Prometheus/Grafana). Also delivered polished React+TypeScript analytics dashboards and designed high-concurrency PostgreSQL schemas with significant latency reductions. Recently built AI-agent orchestration and an interactive node-based requirements dashboard for Siemens Polarion via MCP servers, improving user interaction by ~17.8%+.

Experience

Education

Key Strengths

• Shipped scalable AI-powered recommendation engine in production handling millions of transactions
• Scaled event-driven Python microservices for peak traffic; added Redis caching to improve response times
• Designed secure, versioned FastAPI services using OAuth2 with observability via Prometheus/Grafana
• Built React + TypeScript seller analytics dashboard with Chart.js and Tailwind; improved seller engagement by ~10%
• Designed and optimized PostgreSQL schema for high-concurrency order processing (composite indexes, JSON columns, row-level locking); reduced latency ~50%
• AWS production operations (EC2 Auto Scaling, Lambda, Docker, GitHub Actions CI/CD) and resolved real peak-traffic latency incident via metrics-driven scaling and Kafka workers
• Improved Polarion requirements navigation via interactive node-based dashboard; reported ~17.8%–20% UX/performance improvement
• Addressed recommendation bias toward popular vendors by adding a discovery component to surface low-interaction/new-vendor products
• Shipped full-stack analytics dashboard end-to-end in Next.js (App Router) + TypeScript with server/client component split
• Performance troubleshooting under traffic spikes using Prometheus metrics; improved UX with skeleton screens and query/index optimization
• Postgres schema design at scale (orders/order_items/inventory) with integrity constraints and indexing strategy validated via EXPLAIN ANALYZE
• Built resilient event-driven order-processing workflows with Kafka (idempotency keys, DLQs, circuit breakers) to handle failure modes
• Drove measurable outcomes: +10% seller engagement; ~15% latency reduction after moving to event-driven architecture; reduced post-release bugs ~16%