Retrack is a modern web application that allows users to upload PDF invoices and automatically extracts structured data from them using AI. It breaks down each invoice into individual line items, calculates totals, and generates a clean summary view. Retrack is designed to assist in financial automation workflows and will later serve as a key module within a larger accounting platform (A M A).
Originally developed as part of the AI Challenge Course, Retrack explores the use of LLMs and modern serverless infrastructure for document understanding and automation.
🤖 Key Features:
-
PDF Upload & Parsing: Users can upload invoices in PDF format, which are then analyzed using AI to extract line-by-line items and metadata.
-
AI-Powered Extraction: Instead of relying on template-based OCR or static parsing, Retrack uses large language models to understand the content of invoices, regardless of layout.
-
Summarization: The app generates a clear summary of invoice totals, dates, and vendors, including the breakdown of goods or services.
-
Realtime Workflow Integration: Retrack is designed to be integrated directly into broader systems like accounting platforms, automating invoice ingestion with minimal human input.
⚙️ Tech Stack:
-
Frontend: React + Next.js
-
Backend & DB: Convex
-
Authentication: Clerk
-
AI & Payment Tools:
-
Schematic (Stripe): for structured data handling
-
Inngest: for orchestrating background jobs (invoice processing, retries, etc.)
-
💡 Use Case & Future Integration: Retrack is not a standalone app—it is a foundational tool that will be integrated into AMA, an AI-powered accounting and tax assistant. This modular approach allows for rapid development and clear separation of responsibilities across services.