Automate data extraction & analysis with intelligent document processing
summary
Broadridge automates extraction from SEC filings using Amazon Textract for tables, Amazon Comprehend plus custom models for free‑text, and AWS Step Functions/Lambda for an event‑driven serverless pipeline; it converts thousands of manual hours from a 100+ operations team into machine output that feeds human review and retraining, growing accuracy over time. The system extracts over 5 million data points from ~300,000 meetings, handles quota throttling for parallel Textract calls, and flags cost and MLOps as operational constraints.
AI-written summary · about
chapters (14)
- 0:00Introduction and speakers
- 1:19AWS machine learning stack
- 2:51Document processing challenges
- 7:50AI services for document processing
- 9:15Broadridge Financial overview
- 12:03SEC filings processing challenges
- 14:59Automation goals at Broadridge
- 17:31Partnering with AWS
- 20:27Architecture design
- 22:01Pipeline for table-based data
- 23:53Pipeline for unstructured data
- 25:43Key takeaways for AI projects
- 29:43Next steps and resources
- 30:48Learning and certification
related talks
32mAutomate data extraction & analysis with intelligent document processing
42mExtract information accurately from documents using computer vision
30mIntelligent document processing: Extract data & insights from documents
35mAutomate your document pipeline with intelligent document processing
30mIntelligent document processing for the insurance industry
31mThe Fast and the Curious: Harnessing AI in Financial Data Extraction at Scale
26mEmpower your business use cases
43mImplementing AI Document Processing at Scale: Syngenta