PAPERTRAIL AI
Tech Stack
Next.js, TypeScript, Tailwind CSS, MongoDB, MongoDB Atlas, Mongoose, Firebase Authentication, Google Gemini AI, Gemini Vision, REST APIs, Vercel, Zero-Storage Processing
Description
PaperTrail AI is a full-stack AI-powered document intelligence platform designed to transform unstructured business documents into structured, actionable intelligence.
Users can upload documents such as PDFs and images, which are processed in memory using Google Gemini Vision AI. The system analyzes the document and extracts important business information including obligations, deadlines, payment amounts, contacts, organizations, risks, recommendations, and actionable tasks.
Instead of simply storing uploaded files, PaperTrail AI converts documents into structured intelligence that can be searched, tracked, scheduled, and acted upon through a centralized operations dashboard.
The platform is designed around a privacy-focused zero-storage workflow where original uploaded documents are processed temporarily in memory and discarded after analysis. Only the structured metadata and extracted intelligence required by the application are persisted in MongoDB Atlas.
My Role
- Designed and developed the complete PaperTrail AI full-stack application from architecture and UI design through deployment.
- Built the document upload and in-memory processing workflow for PDF and image-based documents.
- Integrated Google Gemini Vision AI for multimodal document analysis and structured information extraction.
- Designed MongoDB Atlas schemas for documents, tasks, notifications, users, and audit records.
- Implemented Firebase Authentication and protected application routes.
- Built automated extraction of obligations, deadlines, payment information, contacts, organizations, risks, recommendations, and action items.
- Implemented an action-item lifecycle with pending, completed, and archived states.
- Built deadline-aware notifications and an action calendar for extracted obligations.
- Implemented document deletion with cascading removal of linked tasks and notifications.
- Implemented zero-storage document processing so original uploaded files are not permanently stored.
- Built dashboard telemetry showing scanned documents, pending actions, completed actions, and active notifications.
- Added document search, task search, AI-assisted document interaction, and audit-trail functionality.
- Built responsive interfaces for desktop, tablet, and mobile layouts.
- Deployed the production application using Vercel and connected it to MongoDB Atlas and Firebase.
Have a project in mind?
subhashadireddy@gmail.com