AI-Powered Lawsuit Summarization for LELU
Extract structured data from NYPD misconduct complaint PDFs in minutes, not months.
Three-Step Extraction Process
1
Upload PDF
Upload complaint PDFs from NYC Law Department data dumps or NYSCEF.
2
AI Extraction
Vision AI reads each page, extracts narrative summaries and LELU taxonomy fields.
3
Review & Export
Review extractions with source provenance, then export to CSV/TSV for Airtable.
Under the Hood
Two-step AI pipeline using AWS Bedrock
PDF
Complaint document
→
Vision OCR
Qwen3 VL 235B
Reads each page as image
→
Text
Combined OCR output
→
Extraction
Llama 4 Scout
LELU taxonomy mapping
→
Structured Data
Summary + tags + provenance
~3s
per page OCR
~5s
extraction
$0.003
per page
Why Use AI Extraction?
Time Savings
Process 10,000 documents in days instead of months of manual work.
Source Provenance
Every extraction includes the exact source text and page number for verification.
High Recall
AI often finds MORE relevant information than manual review misses.
Low Cost
As low as $0.003/page with cloud APIs, or $0.0002/page with self-hosted OCR.
Quick Cost Estimate
Test Results (5 Cases)
28
Matches
Taxonomy items AI found correctly
22
Misses
Items in ground truth AI missed
+25
Extra Findings
Additional items AI found
$0.003
Cost/Page
Average processing cost