Intelligent Documents Management System
summary
An end-to-end pipeline running on Google Cloud uses Cloud Run, Cloud Build, Bitbucket and Google Cloud Storage to convert scanned files into structured JSON for ERPs, combining object-detection models per document type with the Google Vision API OCR. Stages include loader, splitter, identifier, cleaner, extractor, aggregator, validator and formatter; training typically uses about 200 labeled pages per document type. The system deletes input data for security, supports batch and real‑time APIs, and is already deployed in real customer contexts.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 0:43Why IDM matters
- 1:45From documents to structured data
- 2:21Business needs
- 3:24Architecture
- 6:34Code Structure
- 10:35Implementation
- 11:54Services
- 15:01Splitter
- 16:36Object Identifier
- 17:32Object Transformer
- 19:09Cleaning
- 19:54Extractor
- 21:16Adapters
- 22:21aggregation
- 23:07validation
- 25:55Questions
related talks
29mBetter Understand Your Documents with AI
30mIntelligent document processing: Extract data & insights from documents
42mUsing Deep Learning to Understand Documents
35mAutomate your document pipeline with intelligent document processing
01h 00mComplex Data Ingestion with Open Source AI
08mUnlocking Document Automation with Vertex AI Platform
36mHow to Build a Cloud Native Image Recognition Solution
43mImplementing AI Document Processing at Scale: Syngenta