• Designed and implemented an in-house data extraction API using Tesseract OCR and Flask, eliminating the need for third-party services and saving ₹50,000 daily in document processing costs. The solution automated text extraction with high-precision, replacing previously outsourced workflows.• Developed a deep learning CNN model using TensorFlow that achieved 96.52% accuracy, significantly reducing the processing time for the policy files from 2-3 days to just 3 hours. This model streamlined operations by accelerating large-scale document classification and extraction.• Integrated multiple models built with Keras, TensorFlow, and OpenCV into a unified API, achieving 90.37% accuracy for the automated detection of various national identification documents. The API replaced manual data tagging, boosting efficiency and operational consistency. • Optimized the Aadhar masking process with ResNet Architecture and Pillow, drastically cutting down processing time while improving accuracy, thereby streamlining the document workflow for sensitive data handling.