DocuSense AI

Overview

DocuSense AI is an intelligent document automation solution developed to reduce the manual effort involved in reviewing procurement and tender documents. The system processes unstructured PDF documents, extracts relevant information, validates extracted facts, and generates structured reports through an automated AI pipeline.

Key Features

• Document Processing: Ingests PDF documents and applies text extraction and OCR techniques to handle structured and scanned content
• AI-Powered Extraction: Uses document chunking, embeddings, vector search, and large language models to retrieve and extract relevant information
• Fact Validation: Cross-checks extracted information to identify inconsistencies and potential conflicts before generating results
• Automated Reporting: Converts processed information into structured Word reports, reducing repetitive documentation work
• Interactive Interface: Provides a user-friendly interface for document submission, processing, and reviewing generated results

Architecture

The solution follows a modular AI pipeline consisting of document ingestion, text extraction, chunking, embedding generation, vector storage, retrieval, LLM processing, validation, and report generation. Documents are divided into manageable sections and stored as embeddings in ChromaDB, allowing relevant information to be retrieved during analysis.

AI & Automation

The system combines retrieval-augmented processing with LLM-based analysis to improve how information is extracted from complex documents. Automated validation helps identify conflicting information, while the end-to-end workflow reduces repetitive manual review and reporting tasks.

Workflow

PDF Ingestion → Text Extraction/OCR → Document Chunking → Embeddings → ChromaDB → Retrieval → LLM Analysis → Fact Validation → Word Report Generation

Technologies

Python, Ollama, Llama, ChromaDB, PyMuPDF, OCR, Gradio, Virtual Environment, Retrieval-Augmented Generation (RAG), LLMs, Git