Megha shyam Yellapu

At Bhatia Brothers, I build AI agents for automated cost estimation, reducing manual estimation time by 40% through structured extraction and reasoning pipelines across 1,000+ documents.

I developed LlamaParse OCR pipelines for invoices, contracts, and technical documents, and architected multi-LLM routing across OpenAI, Qwen, and Gemini based on cost, latency, and accuracy. I also build FastAPI services, Dockerized inference workloads, PostgreSQL data pipelines, and AI features for React/Node.js applications on Azure.

Previously at ATOMTE, I fine-tuned Qwen 2.5 on 5,000+ internal documents, reaching 92% response accuracy for domain-specific question answering. I built an end-to-end RAG pipeline that reduced query latency by 35% and deployed a scalable Azure Container Apps API handling 1,000+ daily queries with 99.9% uptime.

I'm also a Smart India Hackathon 2024 finalist and Microsoft Learn Student Ambassador who enjoys building and demonstrating practical full-stack AI solutions.