Document AI
Global Filings PDF Parser (NextGen) – Real-Time Table & Text Extraction
Real-time text, headings, and financial-table extraction from global filings.
Project Overview
Built and operate a Docling-based PDF parser that extracts text, hierarchical headings and financial tables from international corporate filings in real time. A MySQL-driven worker claims queued filings, processes pages in parallel on dedicated servers and stores structured JSON back to the database, using higher-resolution table rendering, automatic repair of fused columns, and operator-supplied or auto-detected tables of contents to set heading levels. Also delivered document reprocessing, latest-first queueing and runtime monitoring.
Tools & Technologies: Python · Docling (TableFormer) · PyTorch · pypdfium2 · Docker · MySQL · multiprocessing · Supervisor · Linux
Parallel page processing
Structured JSON output
Document reprocessing
Complex layout handling
Project capabilities are described from recorded experience. Artwork uses illustrative sample data.
Need a similar system?
Let’s discuss how I can help build a solution around your documents, data, and workflows.
More to explore
All projects →
AWS Announcements Pipeline
Parallel serverless PDF parsing for a stock-exchange announcements feed.
SEC Filings Table Extraction
Semantic financial-table JSON from live SEC filings and the 2006–2026 archive.
Extracted-Table Review Portal
Source-linked PDF comparisons, table review views, and structured-output inspection.