ADVANCED AI-POWERED DOCUMENT INTELLIGENCE SYSTEM

Authors:

Bhanusri¹, K. Dheeraj², V. Harshitha³, P. Chandini⁴, N. Arun Kumar⁵

^{1,2,3,4 5}MVGR College of Engineering, Vizianagaram, India

Abstract—The exponential growth of digital documentation necessitates advanced automated tools for information extraction. Traditional cloud-based solutions, while powerful, pose significant data privacy risks and recurring costs. This paper presents an Advanced AI-Powered Document Intelligence System that utilizes locally deployed Large Language Models (LLMs) and Optical Character Recognition (OCR). By leveraging Llama 3.2 via the Ollama framework and fine-tuning with Low-Rank Adaptation (LoRA), the system achieves enterprise-grade summarization and question-answering capabilities on local hardware. Features include multimodal interaction (TTS/STT), MCQ generation, and high-accuracy extraction from handwritten documents. Performance is validated using ROUGE metrics, demonstrating that local, privacy- centric AI can match cloud-based performance levels.