ABHAY KUMAR

I’m an AI / Machine Learning Engineer who blends ML, LLM workflows, and backend engineering to ship production systems that make teams faster and operations calmer. At Amazon, I built an internal LLM-powered assistant grounded on service documentation to handle onboarding and operational Q&A for new joiners.

I also developed “Sable Orbit,” a Model Context Protocol (MCP) server that evaluates cluster scheduler statistics, BDB sizes, and throttling metrics to determine peak-readiness for events like Prime Day and Black Friday. This enables automated, data-driven scaling decisions across compaction, rebalancing, and repartitioning.

My work goes beyond models: I integrated LLM-powered engineering workflows for code generation, debugging, and operational diagnostics—improving engineering productivity by approximately 30%. I designed and deployed backend automation services in Python and Java to orchestrate large-scale infrastructure workflows, reducing manual operational effort by 70%.

Across multi-region production services (US, EU, Asia-Pacific), I resolved 50+ critical issues using LLM-assisted diagnostics, root cause analysis, and observability improvements, including high-severity Prime Day peak rotations for clusters sustaining 140–150M TPS. Earlier, at SAP Labs, I delivered AI-backed invoice extraction, REST microservices, event-driven processing with Kafka, and reliable persistence using PostgreSQL—validated through CI/CD with testing and regression.