AI analysis of bankruptcy and restructuring cases2025 – 2026
Deep Dive
Investors in distressed assets need to know what is inside a bankruptcy case before they buy credit in it. The case file is a PDF that can pass 10,000 pages. Deep Dive reads it and writes the report.
- Python pipeline in stages: find the relevant pages, transcribe them, extract structured data, merge everything into one report.
- Designed how context moves between stages and which model handles transcription, reasoning and extraction.
- Outputs a written report plus spreadsheets: list of creditors, credit assignments and pending incidents.
- Scrapers for the São Paulo court systems (e-SAJ and DJE) fill in what the PDF does not contain.
- My role
- Developer of the AI workflows, in direct contact with the client
- Stack
- Python, LLM APIs, OpenRouter, PDF processing, Web scraping, Pandas