At National Institute of Technology, Warangal, I integrated six open-source OCR engines through a modular wrapper architecture, enabling each engine to run independently.
I developed an evaluation engine using CER/WER, word-level precision and recall, and entity-field extraction accuracy. I validated it against 973 SROIE-2019 invoices with 100% ground-truth coverage.
I also designed a multi-page PDF pipeline with per-page OCR, bounding-box stitching, weighted confidence aggregation, and Redis-backed job cancellation. I created an in-browser editor so researchers could modify OCR wrapper logic and rerun inference without redeployment.
In my projects, I developed a Node.js and Express API gateway with five rate-limiting algorithms and Redis-backed Lua scripts. I also built SkillSwap, a full-stack platform for peer-to-peer skill exchange with real-time messaging and secure user sessions.
