Hey, I'm Mangesh Dudhgaonkar PatilSenior Backend Engineer

Powered LLM tool orchestration by building a 0→1 MCP (Model Context Protocol) Server in Java (Spring + Java MCP SDK) and Python (FastMCP, chosen for production) – tool registration, JSON-RPC routing, error handling, LLM client integration for agentic workflows.

Served 200,000+ user queries at 4.1/5 satisfaction by shipping a production RAG assistant on AWS Bedrock – owned end-to-end design, embedding pipelines, vector search, prompt engineering.

Designed and built a 0→1 long-running background job framework (authored design doc) for jobs running hours to weeks across 5 DB regions in parallel on dedicated worker instances – decoupled from web/app servers, insulated from deploys (jobs survive rollouts), with cooperative pause/resume, dry-run mode, and region-scoped routing. Powered multi-billion-record MongoDB migrations and Elasticsearch reindex jobs in production.

Cut cross-region MongoDB query latency by 90% by architecting horizontal sharding of 5B+ records via a fault-tolerant, checkpoint-based migration framework with HA guarantees.

Improved API latency by 25% and grew monthly application creation by 50% on the Onshape App Store – built scalable Spring Boot microservices and REST APIs.

Cleared 10+ enterprise escalations and unblocked Reports for 100,000+ users on Spring MVC – owned subsystems end-to-end via retry mechanisms, HTTP client upgrades, and composite index refactoring.