Projects/WebApps

Open source and freeware tools for legal tech, identity infrastructure, and the occasional whimsical experiment.

Open Source Infrastructure

AgentLedger — AI agent audit trail SDK for regulated industries

Legal & Authentication WebApps

Tools for the authentication cliff — when verification becomes scarce, infrastructure matters.

🔬 DocForensics

Document Authenticity Analysis

VirusTotal, but for document forgery. Metadata extraction, compression artifact detection, ELA visualization, font consistency checking, structure, ID, Heuristics, Signature, and ML-based forgery detection. Document fraud is up 244% — this is the first-pass filter.

Metadata analysis (EXIF, PDF info, date paradoxes, creation software)
Error Level Analysis (ELA) visualization
Clone/copy-move detection
PDF/DOC/DOCX structure analysis (hidden layers, JavaScript, edit history)
Roboflow ML integration

GitHub · Live Demo

⚖️ CiteCheck

AI Citation Verification

Verify AI-generated legal citations before they become sanctions. Batch-process briefs and motions, check each citation against CourtListener's federal case database, get confidence scores and warnings. The Avianca problem.

PDF/DOCX/TXT support
CourtListener API integration
Risk assessment scoring
CLI + Web UI

GitHub · Live Demo

⚖️ SPA Vibe Editor

Natural Language Contract Editing

"Vibe coding" for Stock Purchase Agreements. Describe changes in plain English, AI proposes edits with tracked diffs, humans review and accept. Built on ABA Model SPA templates. Multi-model architecture: Claude for drafting and validation, GPT o3-mini for precedent research ("what's market").

Natural language → legal edits → human review
Deal terms form auto-populates template variables
Precedent search ("what's market for escrow %?")
Change log with rationale for each edit
DOCX export with redline track changes
Quick Vibe presets (Buyer-Friendly, Seller-Friendly, Tighten Materiality, Add Carve-outs)

Live Demo

(Desktop only)

📋 Case Intake

AI-Powered Legal Assessment

Four-stage intake workflow: Intake → Timeline → Research → Assessment. Upload case documents, AI extracts key facts and dates, runs legal research, generates preliminary assessment with risk factors and recommended next steps. Built as proof of concept during lunch with a law firm partner.

PDF/DOCX document parsing
Automatic timeline extraction
Legal research integration
Risk assessment scoring
Stage-by-stage workflow with human checkpoints

GitHub · Live Demo

(Desktop only)

Fun Experiments

Weekend projects.

📚 Spine Reader

Bookshelf Personality Analysis

Upload a photo of your bookshelf. Claude Vision reads the spines, enriches with Open Library data, generates reading statistics, and DALL-E creates a tarot-card style portrait of your "reader archetype." Literary personality test meets computer vision.

Claude Vision spine extraction
Open Library metadata enrichment
DALL-E archetype portrait generation
Genre/decade/author/pages/statistics
Output excel file with all titles/statistics

Spine Reader

How These Were Built

These tools were built in collaboration with Claude Code (Anthropic) — not as one-shot prompts, but through extended working sessions where we iterated on architecture, caught edge cases, and refined the details together.

The code hosted on GitHub is free to use, fork, and improve. If you build something interesting with it, I'd love to hear about it. Most can be deployed to Vercel in one click. Each repo includes a .claude/CLAUDE.md file with project context for Claude Code continuation.