At Inside Infotech Solutions, I fine-tuned EfficientNet-B3 for an 18-class document classification engine, achieving 99.86% accuracy. I also built a FastAPI inference system with confidence-based fallback.
I developed a Gujarati voter list OCR pipeline using the Gemini API to extract voter information from PDFs, then processed and stored the structured data in MSSQL.
My projects include a Graph RAG system using Neo4j knowledge graphs and FAISS vector search, and Briefly, an AI document processing system built with FastAPI, Tesseract OCR, and Llama-3.1. I was also a member of the Top Winning Team at the IBM Expert Labs National Hackathon 2026.

