Sharayu Garad

I’m an AI Engineer and Data Scientist with 2+ years of experience building production LLM systems for real-world legal environments. I focus on end-to-end reliability—designing RAG retrieval pipelines, automating LLM workflows, and making outputs source-traceable through metadata-driven approaches.

At Levi and Korsinsky, LLP, I promoted from Python Developer to Data Scientist, AI Engineer (Jun 2025 to Present), where I architected a production RAG pipeline for legal document analysis using LangChain and FAISS/Pinecone. This work achieved a 30% reduction in hallucinated citations by using source-traceable, metadata-driven retrieval. I also built a two-pass LLM triage pipeline (OpenAI GPT-4o) that classifies 100,000+ Zoho Desk tickets/week into 11 categories with auto-generated sub-categories, tone analysis, escalation flags, and human-review draft replies.

I’ve deployed Claude MCP integrations (ZohoMCP, Meta Ads Connector, PDF Tools MCP) into production workflows, reducing manual ticket routing effort by 95%. I received an Excellence Award for eliminating 99% of manual effort across 100K+ documents, and I bring that same mindset to everything I build—LLM caching, confidence scoring, hallucination prevention, and evaluation practices (including human-in-the-loop) so teams can trust the automation.