I manage courtroom operations at Santa Barbara Superior Court — nine years of watching systems work, fail, and get blamed on the wrong thing.
I started building AI tools in July 2025. I use AI as a thinking partner, move fast, and ship real things. Growth is measurable. Context shouldn't evaporate. Invisible problems deserve visible solutions.
I'm also a Supernote devotee, a Lakers lifer, and firmly convinced that earl grey hot is the correct choice.
📂 Portfolio → beyond-the-docket-489960083310.us-west1.run.app
Named for the thread Ariadne gave Theseus to navigate the labyrinth — and find his way back.
A suite of tools for AI-native builders who move fast and need infrastructure that keeps up. Each tool solves one problem completely. They're designed to work together, but they don't have to.
🪬 Janus Clew — Developer Growth Tracking
You've shipped three projects in three months. You're definitely better — but can you prove it?
Janus reads your actual code, measures complexity objectively, and exports proof you can share. Not vibes. Not self-reporting. Evidence.
🏆 AWS Global Vibe Hackathon 2025
🧵 Ariadne Clew — Reasoning Preservation
You build fast. You iterate messy. You lose track of why you made choices.
Ariadne turns your chaotic LLM chat sessions into structured clarity — capturing the aha moments, design tradeoffs, and code decisions that disappear when you move too fast for version control. Works with ChatGPT, Claude, DeepSeek, whatever you use.
Real AgentCore integration. No mocks.
🔮 Metis Clew — AI Code Explainer
When you're learning to code, you hit this wall: code doesn't feel clear. You paste snippets into AI tools and get long paragraphs that don't match what you're actually trying to understand.
Metis shows you WHAT it does, HOW it works, and WHY it was built that way. Validated through user research with career switchers. MVP shipped in 5 weeks.
🧠 Memoria Clew — Research Memory
You save a link while debugging a feature. Later, the link is still there — but the reason you saved it is gone. The context evaporates.
Bookmarks don't remember why. Memoria does. Local-first research memory that resurfaces what you saved when it becomes relevant again, with transparent reasoning. No black-box magic.
AI Vibe Coding Hackathon 2026 · 4th place / 250 projects
🧭 Clew Directive — AI Learning Paths
AI education is overwhelming — thousands of courses, most behind paywalls, few personalized.
4 questions. A personalized learning path from verified free resources. A PDF you keep forever. No signup. No data stored. Multi-agent system (Scout + Navigator + Curator) on AWS serverless. Free, private, accessible to all.
🪶 Hermes Clew — Agent Readiness Scanner
The web is about to get a second audience. AI agents can't use most web apps — they can't find buttons, fill forms, or navigate. They see a wall of <div onclick="..."> and bounce.
Hermes scans your HTML/JSX/TSX files and tells you exactly what's broken and how to fix it, scored 0-100. Lives in GitLab Duo Chat. Just ask it to scan.
GitLab Duo Agent Platform Challenge 2026
🛡️ Anton Clew — AI Workflow Safety Guardrails
AI coding tools are powerful — but they are not inherently safe. Automation can modify sensitive files unintentionally, change critical configuration without context, or push large batches of code faster than meaningful human review can happen.
Anton is a policy-driven CLI that classifies file changes as DENY, NEEDS_APPROVAL, or ALLOW before they land. Deterministic rules. No probabilistic guesses. No hallucinations.
🎫 TicketGlass — IT Support Transparency (Iris Clew)
IT support tickets are a black box. Customers don't know what's happening, why it's happening, or if the same issue will happen again next month.
Real-time visibility, AI-powered explanations, smart solution tracking.
Built for SuperHack 2025
🕯️ Lumen Clew — Code Health Scanner
Most code feedback comes with judgment baked in. Lumen doesn't do that.
Paste any public GitHub repo URL and get insights across code quality, dependency health, secret detection, and accessibility — with Claude translating findings into plain English. Awareness, not shame. Reflection, not prescription.
Live app → lumenclew.lovable.app
🔧 Clew Forge — Project Framework
Where your reasoning threads become real artifacts. Lightweight framework for builders shipping fast without losing clarity. Start with 3 files.
The work I do at the intersection of courts, data, and people who deserve better tools.
| Project | What it does |
|---|---|
| Jury Eligibility Chatbot | Answers jury service questions using CA Code of Civil Procedure + RAG on Bedrock |
| 12 Angry Rows | 11 years of jury utilization data, benchmarked against NCSC standards |
| Write of Access | AWS Textract on public court forms — and honest documentation of what it can't do |
| Hung Jury | Explored automating Crystal Reports PDFs. The blocker was data quality, not Python. Sometimes the finding is the deliverable. |
| Panoptes Scout | First-pass data quality briefing agent for tabular datasets — Google AI Agents capstone |
Sprint-Kit — Teaching middle schoolers real PM skills. AI scaffolds task breakdown. Zero data collection.
Proof of Scoville 🌶️ — On-chain salsa tracker built for Proof of Concept 2025 Hackathon. Yes, really. No regrets.
Office Keys — Gamified typing practice for workforce digital literacy, because some people are still getting left behind by something as basic as a keyboard.
Bench Power: Lakers 🏀 — Data storytelling on bench impact. Google Analytics Capstone. Lakers forever, that's non-negotiable.
| What | Details | |
|---|---|---|
| 🥇 | DoraHacks AWS Hackathon (2025) | Prize winner — Draper University scholarship · Lovable access · AWS credits |
| 🎤 | NACM 2026 Workshop (Accepted) | "Building Bridges: How Court Leaders Can Enable Staff Innovation" |
| 🏅 | AI Vibe Coding Hackathon (2026) | 4th place out of 250 projects |
| 🥈 | CCC AI Summer Camp (2025) | 2nd place out of 14 teams |
Occam's Razor — Solve one problem completely before adding the next. Simplest solution that ships bulletproof scores higher than ambitious solution that crashes.
Single Responsibility — Every component does one thing well. No Swiss Army knives.
Security-First — PII protection, content filtering, guardrails baked in early. Not bolted on after.
Foundation-First — Real integrations over mocked features. Production architecture over quick hacks. This is always a v1, not a v-final.
Operational Thinking — Nine years in judicial systems means I know implementations have real-world consequences. That instinct doesn't turn off.
📍 Santa Barbara, CA · 🖋️ Supernote devotee · 🏀 Lakers forever
AI-assisted development. Human reviewed. All mistakes are mine.