BOUCH Legal

Legal & Evidence Tools

Evidence analysis toolkit. 1,300+ items processed: documents, emails, 850+ images. 9 forensic reports per run, 644 cross-evidence correlations. Built for a real employment law case.

The Story

Real employment law case, real stakes. 1,300+ evidence items across documents, emails, and 850+ photographs. 4-stage pipeline: ingest (SHA256 content-addressed storage), analyse (GPT-4o Vision + text), correlate (cross-evidence pattern matching), package (9 forensic reports per run).

6 generations, each building on the last. Started with basic document analysis. Added image OCR, entity correlation, forensic integrity with chain of custody, 48 Pydantic models, and AI entity resolution (+114% yield). Same pipeline works in compliance, insurance, HR, investigations.

Stage 1 Ingest
Stage 2 Analyze
Stage 3 Correlate
Stage 4 Report

6 Generations of Capability

Each generation compounds on the last — not rewrites, each new capability adds to the existing pipeline.

Gen 1
Document analysis Timeline reconstruction Basic entity extraction
Gen 2
Document analysis Timeline Image OCR + forensics Vision AI (GPT-4o)
Gen 3
Docs OCR Timeline Entity correlation Cross-evidence matching Legal pattern detection
Gen 4
All previous SHA256 evidence management Content-addressed storage Chain of custody
Gen 5
All previous 48 Pydantic models Template Method reports Structured validation
Gen 6
All previous AI entity resolution Smart defaults +114% entity yield

Tools

Evidence Toolkit v4.1

In Development

Core forensic analysis platform. 1,300+ items processed, 644 cross-evidence correlations, 9 report types (executive summary, financial risk, timeline, legal patterns, entity network, power dynamics, OCR, quoted statements, forensic opinion). SHA256 chain of custody, 48 Pydantic models, content-addressed storage.

PythonOpenAIPydanticSHA256

Document Evidence Analyzer

In Development

Text analysis for legal evidence. Extracts key facts, identifies contradictions, builds timelines, and highlights patterns across large document sets.

NLPDocument IntelligencePython

Image Evidence Analyzer

In Development

GPT-4o Vision forensic analysis. 850+ images processed with 0.89 average confidence. OCR text extraction, scene description, object detection, legal relevance scoring across 6 domains. Thread-safe parallel processing at ~$0.001/image.

GPT-4o VisionOCRForensicsPython

Legal AI System

In Development

Legal Document Intelligence and Decision Support System. Built with FastAPI, PydanticAI, Supabase, and MCP. Expert knowledge base with scoping, multi-agent analysis, and structured legal reasoning.

FastAPIPydanticAIMCPSupabase

Approach

Domain-Transferable

Same evidence analysis patterns work in compliance, insurance, HR, investigations. The toolkit adapts to any document-heavy domain.

AI-Native

Built with PydanticAI and MCP from the ground up. Structured outputs, typed tool calls, reproducible analysis chains.

Battle-Tested

Used in a live case. Real document volumes, real evidence standards, real deadlines.

Domain-Transferable

Same pipeline works in compliance, insurance, HR investigations, regulatory cases. Different source material, same method.

Get in touch →

Free Your Main Thread