Overview
A financial research team at Brock's Goodman School of Business needed to analyze executive compensation data buried inside 25,000+ SEC regulatory filings spanning 2017–2024. The existing process was entirely manual.
The Challenge
Manually reading, classifying, and extracting structured data from thousands of complex legal/financial documents is impossibly slow. The team needed a system that could do this at scale with a high degree of accuracy — and keep running reliably under API rate limits and quota constraints.
What Was Built
- Automated data pipeline to fetch, parse, and process 25,000+ SEC filings from EDGAR
- Multi-model AI classification workflow using OpenAI, Claude, and Gemini APIs in parallel for cross-validation
- Batch processing system with resume capabilities so large runs could be paused and continued
- Rate limiting and quota management layer to handle API constraints across three providers
- Structured output extraction with >95% classification accuracy on executive compensation data